2026-03-06 20:57:27 +08:00
{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## 1. 导入依赖"
]
},
{
"cell_type": "code",
"metadata": {
"ExecuteTime": {
2026-03-09 23:37:20 +08:00
"end_time": "2026-03-09T15:26:35.493553Z",
"start_time": "2026-03-09T15:26:34.919597Z"
2026-03-06 20:57:27 +08:00
}
},
"source": [
"import os\n",
"from datetime import datetime\n",
"from typing import List\n",
"\n",
"import polars as pl\n",
"\n",
"from src.factors import FactorEngine\n",
"from src.training import (\n",
" DateSplitter,\n",
" LightGBMModel,\n",
" STFilter,\n",
" StandardScaler,\n",
2026-03-09 22:33:41 +08:00
" # StockFilterConfig, # 已删除，使用 StockPoolManager + filter_func 替代\n",
2026-03-06 20:57:27 +08:00
" StockPoolManager,\n",
" Trainer,\n",
" Winsorizer,\n",
" NullFiller,\n",
")\n",
"from src.training.config import TrainingConfig"
],
"outputs": [],
2026-03-09 22:33:41 +08:00
"execution_count": 1
2026-03-06 20:57:27 +08:00
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## 2. 定义辅助函数"
]
},
{
"metadata": {
"ExecuteTime": {
2026-03-09 23:37:20 +08:00
"end_time": "2026-03-09T15:26:35.507113Z",
"start_time": "2026-03-09T15:26:35.502592Z"
2026-03-06 20:57:27 +08:00
}
},
"cell_type": "code",
"source": [
"def create_factors_with_strings(engine: FactorEngine, factor_definitions: dict, label_factor: dict) -> List[str]:\n",
" print(\"=\" * 80)\n",
" print(\"使用字符串表达式定义因子\")\n",
" print(\"=\" * 80)\n",
"\n",
" # 注册所有特征因子\n",
" print(\"\\n注册特征因子:\")\n",
" for name, expr in factor_definitions.items():\n",
" engine.add_factor(name, expr)\n",
" print(f\" - {name}: {expr}\")\n",
"\n",
" # 注册 label 因子\n",
" print(\"\\n注册 Label 因子:\")\n",
" for name, expr in label_factor.items():\n",
" engine.add_factor(name, expr)\n",
" print(f\" - {name}: {expr}\")\n",
"\n",
" # 从字典自动获取特征列\n",
" feature_cols = list(factor_definitions.keys())\n",
"\n",
" print(f\"\\n特征因子数: {len(feature_cols)}\")\n",
" print(f\"Label: {list(label_factor.keys())[0]}\")\n",
" print(f\"已注册因子总数: {len(engine.list_registered())}\")\n",
"\n",
" return feature_cols\n",
"\n",
"\n",
"def prepare_data(\n",
2026-03-09 22:33:41 +08:00
" engine: FactorEngine,\n",
" feature_cols: List[str],\n",
" start_date: str,\n",
" end_date: str,\n",
2026-03-06 20:57:27 +08:00
") -> pl.DataFrame:\n",
" print(\"\\n\" + \"=\" * 80)\n",
" print(\"准备数据\")\n",
" print(\"=\" * 80)\n",
"\n",
" # 计算因子（全市场数据）\n",
" print(f\"\\n计算因子: {start_date} - {end_date}\")\n",
2026-03-08 14:12:03 +08:00
" factor_names = feature_cols + [LABEL_NAME] # 包含 label\n",
2026-03-06 20:57:27 +08:00
"\n",
" data = engine.compute(\n",
" factor_names=factor_names,\n",
" start_date=start_date,\n",
" end_date=end_date,\n",
" )\n",
"\n",
" print(f\"数据形状: {data.shape}\")\n",
" print(f\"数据列: {data.columns}\")\n",
" print(f\"\\n前5行预览:\")\n",
" print(data.head())\n",
"\n",
" return data"
],
"outputs": [],
2026-03-09 22:33:41 +08:00
"execution_count": 2
2026-03-06 20:57:27 +08:00
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## 3. 配置参数\n",
"\n",
"### 3.1 因子定义"
]
},
{
"metadata": {
"ExecuteTime": {
2026-03-09 23:37:20 +08:00
"end_time": "2026-03-09T15:26:35.516098Z",
"start_time": "2026-03-09T15:26:35.511343Z"
2026-03-06 20:57:27 +08:00
}
},
"cell_type": "code",
"source": [
"# 特征因子定义字典：新增因子只需在此处添加一行\n",
2026-03-08 14:12:03 +08:00
"LABEL_NAME = 'future_return_5'\n",
"\n",
2026-03-09 23:37:20 +08:00
"FACTOR_DEFINITIONS = FACTOR_DICT = {\n",
" # ================= 1. 价格、趋势与路径依赖 (Trend, Momentum & Path Dependency) =================\n",
2026-03-08 23:14:18 +08:00
" \"ma_5\": \"ts_mean(close, 5)\",\n",
" \"ma_20\": \"ts_mean(close, 20)\",\n",
2026-03-09 23:37:20 +08:00
" \"ma_ratio_5_20\": \"ts_mean(close, 5) / (ts_mean(close, 20) + 1e-8) - 1\", # 均线发散度\n",
" \"bias_10\": \"close / (ts_mean(close, 10) + 1e-8) - 1\", # 10日乖离率\n",
" \"high_low_ratio\": \"(close - ts_min(low, 20)) / (ts_max(high, 20) - ts_min(low, 20) + 1e-8)\", # 威廉指标变形\n",
" \"bbi_ratio\": \"(ts_mean(close, 3) + ts_mean(close, 6) + ts_mean(close, 12) + ts_mean(close, 24)) / (4 * close + 1e-8)\", # 多空指标比率\n",
" \"return_5\": \"(close / (ts_delay(close, 5) + 1e-8)) - 1\", # 5日动量\n",
" \"return_20\": \"(close / (ts_delay(close, 20) + 1e-8)) - 1\", # 20日动量\n",
"\n",
" # [高阶] Kaufman 趋势效率 (极高价值) - 衡量趋势流畅度，剔除无序震荡\n",
" \"kaufman_ER_20\": \"abs(close - ts_delay(close, 20)) / (ts_sum(abs(close - ts_delay(close, 1)), 20) + 1e-8)\",\n",
" # [高阶] 动量加速度 - 寻找二阶导数大于0，
" \"mom_acceleration_10_20\": \"(close / (ts_delay(close, 10) + 1e-8) - 1) - (ts_delay(close, 10) / (ts_delay(close, 20) + 1e-8) - 1)\",\n",
" #[高阶] 高点距离衰减 - 衡量套牢盘压力\n",
" \"drawdown_from_high_60\": \"close / (ts_max(high, 60) + 1e-8) - 1\",\n",
" # [高阶] 趋势一致性 - 过去20天内收红的天数比例\n",
" \"up_days_ratio_20\": \"ts_sum(close > ts_delay(close, 1), 20) / 20\",\n",
"\n",
" # ================= 2. 波动率、风险调整与高阶矩 (Volatility & Risk-Adjusted Returns) =================\n",
" \"volatility_5\": \"ts_std(close, 5)\",\n",
" \"volatility_20\": \"ts_std(close, 20)\",\n",
" \"volatility_ratio\": \"ts_std(close, 5) / (ts_std(close, 20) + 1e-8)\", # 波动率期限结构\n",
" \"std_return_20\": \"ts_std((close / (ts_delay(close, 1) + 1e-8)) - 1, 20)\", # 真实收益率波动率\n",
"\n",
" # [高阶] 夏普趋势比率 - 惩罚暴涨暴跌，奖励稳健爬坡\n",
" \"sharpe_ratio_20\": \"ts_mean(close / (ts_delay(close, 1) + 1e-8) - 1, 20) / (ts_std(close / (ts_delay(close, 1) + 1e-8) - 1, 20) + 1e-8)\",\n",
" #[高阶] 尾部崩盘风险 - 过去一个月最大单日跌幅\n",
" \"min_ret_20\": \"ts_min(close / (ts_delay(close, 1) + 1e-8) - 1, 20)\",\n",
" # [高阶] 波动率挤压比 - 寻找盘整到极致面临变盘的股票 (布林带收口)\n",
" \"volatility_squeeze_5_60\": \"ts_std(close, 5) / (ts_std(close, 60) + 1e-8)\",\n",
"\n",
" # ================= 3. 日内微观结构与异象 (Intraday Microstructure & Anomalies) =================\n",
" # [高阶] 隔夜与日内背离 - 差值越小说明主力越喜欢在盘中吸筹\n",
" \"overnight_intraday_diff\": \"(open / (ts_delay(close, 1) + 1e-8) - 1) - (close / (open + 1e-8) - 1)\",\n",
" #[高阶] 上影线抛压极值 - 冲高回落被套牢的概率\n",
" \"upper_shadow_ratio\": \"(high - ((open + close + abs(open - close)) / 2)) / (high - low + 1e-8)\",\n",
" # [高阶] 资金沉淀率 - 衡量主力日内高抛低吸洗盘的剧烈程度\n",
" \"capital_retention_20\": \"ts_sum(abs(close - open), 20) / (ts_sum(high - low, 20) + 1e-8)\",\n",
" # [高阶] MAX 彩票效应 - 反转因子，剔除近期有过妖股连板特征的标的\n",
" \"max_ret_20\": \"ts_max(close / (ts_delay(close, 1) + 1e-8) - 1, 20)\",\n",
"\n",
" # ================= 4. 量能、流动性与量价背离 (Volume, Liquidity & Divergence) =================\n",
" \"volume_ratio_5_20\": \"ts_mean(vol, 5) / (ts_mean(vol, 20) + 1e-8)\", # 相对放量比\n",
" \"turnover_rate_mean_5\": \"ts_mean(turnover_rate, 5)\", # 活跃度\n",
" \"turnover_deviation\": \"(turnover_rate - ts_mean(turnover_rate, 10)) / (ts_std(turnover_rate, 10) + 1e-8)\", # 换手率偏离度\n",
"\n",
" # [高阶] Amihud 非流动性异象 (绝对核心) - 衡量砸盘/拉升的摩擦成本\n",
" \"amihud_illiq_20\": \"ts_mean(abs(close / (ts_delay(close, 1) + 1e-8) - 1) / (amount + 1e-8), 20)\",\n",
" # [高阶] 换手率惩罚因子 - 换手率忽高忽低说明游资接力，行情极不稳定\n",
" \"turnover_cv_20\": \"ts_std(turnover_rate, 20) / (ts_mean(turnover_rate, 20) + 1e-8)\",\n",
" # [高阶] 纯粹量价相关性 - 检验是否是\"放量上涨，缩量下跌\"的良性多头\n",
" \"pv_corr_20\": \"ts_corr(close / (ts_delay(close, 1) + 1e-8) - 1, vol, 20)\",\n",
" # [高阶] 收盘价与均价背离 - 专门抓尾盘突袭拉升骗线的股票\n",
" \"close_vwap_deviation\": \"close / (amount / (vol * 100 + 1e-8) + 1e-8) - 1\",\n",
"\n",
" # ================= 5. 基本面财务特征 (Fundamental Quality & Structure) =================\n",
" \"roe\": \"n_income / (total_hldr_eqy_exc_min_int + 1e-8)\", # 净资产收益率\n",
" \"roa\": \"n_income / (total_assets + 1e-8)\", # 总资产收益率\n",
" \"profit_margin\": \"n_income / (revenue + 1e-8)\", # 销售净利率\n",
" \"debt_to_equity\": \"total_liab / (total_hldr_eqy_exc_min_int + 1e-8)\", # 杠杆率\n",
" \"current_ratio\": \"total_cur_assets / (total_cur_liab + 1e-8)\", # 短期偿债安全垫\n",
"\n",
" #[高阶] 利润同比增速 (日频延后252天等于去年同期)\n",
" \"net_profit_yoy\": \"(n_income / (ts_delay(n_income, 252) + 1e-8)) - 1\",\n",
" # [高阶] 营收同比增速\n",
" \"revenue_yoy\": \"(revenue / (ts_delay(revenue, 252) + 1e-8)) - 1\",\n",
" # [高阶] 资产负债表扩张斜率 - 剔除单纯靠举债扩张的公司\n",
" \"healthy_expansion_velocity\": \"(total_assets / (ts_delay(total_assets, 252) + 1e-8) - 1) - (total_liab / (ts_delay(total_liab, 252) + 1e-8) - 1)\",\n",
"\n",
" # ================= 6. 基本面估值与截面动量共振 (Valuation & Cross-Sectional Ranking) =================\n",
" # 估值水平绝对值 (Tushare 市值单位需要 * 10000 转换为元)\n",
" \"EP\": \"n_income / (total_mv * 10000 + 1e-8)\", # 盈利收益率 (1/PE)\n",
" \"BP\": \"total_hldr_eqy_exc_min_int / (total_mv * 10000 + 1e-8)\", # 账面市值比 (1/PB)\n",
" \"CP\": \"n_cashflow_act / (total_mv * 10000 + 1e-8)\", # 经营现金流收益率 (1/PCF)\n",
"\n",
" # 全市场截面排名因子\n",
" \"market_cap_rank\": \"cs_rank(total_mv)\", # 规模因子 (Size)\n",
" \"turnover_rank\": \"cs_rank(turnover_rate)\",\n",
" \"return_5_rank\": \"cs_rank((close / (ts_delay(close, 5) + 1e-8)) - 1)\",\n",
" \"EP_rank\": \"cs_rank(n_income / (total_mv + 1e-8))\", # 谁最便宜\n",
"\n",
" # [高阶] 戴维斯双击动量 - 估值相对上一年是否在扩张\n",
" \"pe_expansion_trend\": \"(total_mv / (n_income + 1e-8)) / (ts_delay(total_mv, 60) / (ts_delay(n_income, 60) + 1e-8) + 1e-8) - 1\",\n",
" # [高阶] 业绩与价格背离度 - 截面做差： ， ，
" \"value_price_divergence\": \"cs_rank((n_income - ts_delay(n_income, 252)) / (abs(ts_delay(n_income, 252)) + 1e-8)) - cs_rank(close / (ts_delay(close, 20) + 1e-8))\",\n",
" # [高阶] 流动性溢价调整后市值 - 识别僵尸大盘股和极度活跃的小微盘\n",
" \"active_market_cap\": \"total_mv * ts_mean(turnover_rate, 20)\",\n",
" \"ebit_rank\": \"cs_rank(ebit)\",\n",
2026-03-06 20:57:27 +08:00
"}\n",
"\n",
"# Label 因子定义（不参与训练，用于计算目标）\n",
"LABEL_FACTOR = {\n",
2026-03-09 22:33:41 +08:00
" LABEL_NAME: \"(ts_delay(close, -5) / ts_delay(open, -1)) - 1\", # 未来5日收益率\n",
2026-03-06 20:57:27 +08:00
"}"
],
"outputs": [],
2026-03-09 22:33:41 +08:00
"execution_count": 3
2026-03-06 20:57:27 +08:00
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### 3.2 训练参数配置"
]
},
{
"cell_type": "code",
"metadata": {
"ExecuteTime": {
2026-03-09 23:37:20 +08:00
"end_time": "2026-03-09T15:26:35.526730Z",
"start_time": "2026-03-09T15:26:35.522343Z"
2026-03-06 20:57:27 +08:00
}
},
"source": [
2026-03-08 01:09:47 +08:00
"# 日期范围配置（正确的 train/val/test 三分法）\n",
"# Train: 用于训练模型参数\n",
"# Val: 用于验证/早停/调参（位于 train 之后，
"# Test: 仅用于最终评估，完全独立于训练过程\n",
2026-03-06 20:57:27 +08:00
"TRAIN_START = \"20200101\"\n",
2026-03-08 01:09:47 +08:00
"TRAIN_END = \"20231231\"\n",
"VAL_START = \"20240101\"\n",
"VAL_END = \"20241231\"\n",
2026-03-06 20:57:27 +08:00
"TEST_START = \"20250101\"\n",
2026-03-09 22:33:41 +08:00
"TEST_END = \"20261231\"\n",
2026-03-06 20:57:27 +08:00
"\n",
"# 模型参数配置\n",
"MODEL_PARAMS = {\n",
" \"objective\": \"regression\",\n",
" \"metric\": \"mae\", # 改为 MAE，
" # 树结构控制（防过拟合核心）\n",
" \"num_leaves\": 20, # 从31降为20，
2026-03-09 22:33:41 +08:00
" \"max_depth\": 5, # 显式限制深度，防止过度拟合噪声\n",
2026-03-06 20:57:27 +08:00
" \"min_child_samples\": 50, # 叶子最小样本数，防止学习极端样本\n",
" \"min_child_weight\": 0.001,\n",
" # 学习参数\n",
" \"learning_rate\": 0.01, # 降低学习率，配合更多树\n",
" \"n_estimators\": 1000, # 增加树数量，配合早停\n",
" # 采样策略（关键防过拟合）\n",
" \"subsample\": 0.8, # 每棵树随机采样80%数据（行采样）\n",
" \"subsample_freq\": 5, # 每5轮迭代进行一次 subsample\n",
" \"colsample_bytree\": 0.8, # 每棵树随机选择80%特征（列采样）\n",
" # 正则化\n",
" \"reg_alpha\": 0.1, # L1正则，
" \"reg_lambda\": 1.0, # L2正则，
" # 数值稳定性\n",
" \"verbose\": -1,\n",
" \"random_state\": 42,\n",
"}\n",
"\n",
"# 数据处理器配置\n",
"PROCESSOR_CONFIGS = [\n",
" {\"name\": \"winsorizer\", \"params\": {\"lower\": 0.01, \"upper\": 0.99}},\n",
" {\"name\": \"cs_standard_scaler\", \"params\": {}},\n",
"]\n",
"\n",
2026-03-09 22:33:41 +08:00
"\n",
"# 股票池筛选函数\n",
"# 使用新的 StockPoolManager API：
"# 筛选函数接收单日 DataFrame，
"#\n",
"# 筛选逻辑（针对单日数据）：\n",
"# 1. 先排除创业板、科创板、北交所（ ）
"# 2. 然后选取市值最小的500只股票\n",
"def stock_pool_filter(df: pl.DataFrame) -> pl.Series:\n",
" \"\"\"股票池筛选函数（单日数据）\n",
"\n",
" 筛选条件：\n",
" 1. 排除创业板（代码以 300 开头）\n",
" 2. 排除科创板（代码以 688 开头）\n",
" 3. 排除北交所（代码以 8、9 或 4 开头）\n",
" 4. 选取当日市值最小的500只股票\n",
" \"\"\"\n",
" # 代码筛选（排除创业板、科创板、北交所）\n",
" code_filter = (\n",
" ~df[\"ts_code\"].str.starts_with(\"300\") & # 排除创业板\n",
" ~df[\"ts_code\"].str.starts_with(\"688\") & # 排除科创板\n",
" ~df[\"ts_code\"].str.starts_with(\"8\") & # 排除北交所\n",
" ~df[\"ts_code\"].str.starts_with(\"9\") & # 排除北交所\n",
" ~df[\"ts_code\"].str.starts_with(\"4\") # 排除北交所\n",
" )\n",
"\n",
" # 在已筛选的股票中，
" # 按市值升序排序，
" valid_df = df.filter(code_filter)\n",
" n = min(1000, len(valid_df))\n",
" small_cap_codes = valid_df.sort(\"total_mv\").head(n)[\"ts_code\"]\n",
"\n",
" # 返回布尔 Series：
" return df[\"ts_code\"].is_in(small_cap_codes)\n",
"\n",
"\n",
"# 定义筛选所需的基础列\n",
"STOCK_FILTER_REQUIRED_COLUMNS = [\"total_mv\"] # ST过滤由STFilter组件处理\n",
"\n",
"# 可选：定义筛选所需的因子（如果需要用因子进行筛选）\n",
"# STOCK_FILTER_REQUIRED_FACTORS = {\n",
"# \"market_cap_rank\": \"cs_rank(total_mv)\",\n",
"# }\n",
"\n",
2026-03-06 20:57:27 +08:00
"\n",
"# 输出配置（相对于本文件所在目录）\n",
"OUTPUT_DIR = \"output\"\n",
"SAVE_PREDICTIONS = True\n",
2026-03-08 11:46:30 +08:00
"PERSIST_MODEL = False\n",
"\n",
"# Top N 配置：每日推荐股票数量\n",
2026-03-08 20:58:35 +08:00
"TOP_N = 5 # 可调整为 10, 20 等"
2026-03-06 20:57:27 +08:00
],
"outputs": [],
2026-03-09 22:33:41 +08:00
"execution_count": 4
2026-03-06 20:57:27 +08:00
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## 4. 训练流程\n",
"\n",
"### 4.1 初始化组件"
]
},
{
"metadata": {
"ExecuteTime": {
2026-03-09 23:37:20 +08:00
"end_time": "2026-03-09T15:26:58.293090Z",
"start_time": "2026-03-09T15:26:35.532653Z"
2026-03-06 20:57:27 +08:00
}
},
2026-03-09 22:33:41 +08:00
"cell_type": "code",
2026-03-06 20:57:27 +08:00
"source": [
"print(\"\\n\" + \"=\" * 80)\n",
"print(\"LightGBM 回归模型训练\")\n",
"print(\"=\" * 80)\n",
"\n",
"# 1. 创建 FactorEngine\n",
"print(\"\\n[1] 创建 FactorEngine\")\n",
"engine = FactorEngine()\n",
"\n",
"# 2. 使用字符串表达式定义因子\n",
"print(\"\\n[2] 定义因子（字符串表达式）\")\n",
"feature_cols = create_factors_with_strings(engine, FACTOR_DEFINITIONS, LABEL_FACTOR)\n",
2026-03-08 14:12:03 +08:00
"target_col = LABEL_NAME\n",
2026-03-06 20:57:27 +08:00
"\n",
"# 3. 准备数据（使用模块级别的日期配置）\n",
"print(\"\\n[3] 准备数据\")\n",
"\n",
"data = prepare_data(\n",
" engine=engine,\n",
" feature_cols=feature_cols,\n",
" start_date=TRAIN_START,\n",
" end_date=TEST_END,\n",
")\n",
"\n",
"# 4. 打印配置信息\n",
"print(f\"\\n[配置] 训练期: {TRAIN_START} - {TRAIN_END}\")\n",
2026-03-08 01:09:47 +08:00
"print(f\"[配置] 验证期: {VAL_START} - {VAL_END}\")\n",
2026-03-06 20:57:27 +08:00
"print(f\"[配置] 测试期: {TEST_START} - {TEST_END}\")\n",
"print(f\"[配置] 特征数: {len(feature_cols)}\")\n",
"print(f\"[配置] 目标变量: {target_col}\")\n",
"\n",
"# 5. 创建模型\n",
"model = LightGBMModel(params=MODEL_PARAMS)\n",
"\n",
"# 6. 创建数据处理器\n",
"processors = [\n",
" NullFiller(strategy=\"mean\"),\n",
" Winsorizer(**PROCESSOR_CONFIGS[0][\"params\"]),\n",
" StandardScaler(exclude_cols=[\"ts_code\", \"trade_date\", target_col]),\n",
"]\n",
"\n",
2026-03-08 01:09:47 +08:00
"# 7. 创建数据划分器（正确的 train/val/test 三分法）\n",
"# Train: 训练模型参数 | Val: 验证/早停 | Test: 最终评估\n",
2026-03-06 20:57:27 +08:00
"splitter = DateSplitter(\n",
" train_start=TRAIN_START,\n",
" train_end=TRAIN_END,\n",
2026-03-08 01:09:47 +08:00
" val_start=VAL_START,\n",
" val_end=VAL_END,\n",
2026-03-06 20:57:27 +08:00
" test_start=TEST_START,\n",
" test_end=TEST_END,\n",
")\n",
"\n",
"# 8. 创建股票池管理器\n",
2026-03-09 22:33:41 +08:00
"# 使用新的 API：
2026-03-06 20:57:27 +08:00
"pool_manager = StockPoolManager(\n",
2026-03-09 22:33:41 +08:00
" filter_func=stock_pool_filter,\n",
" required_columns=STOCK_FILTER_REQUIRED_COLUMNS, # 筛选所需的额外列\n",
" # required_factors=STOCK_FILTER_REQUIRED_FACTORS, # 可选：筛选所需的因子\n",
2026-03-06 20:57:27 +08:00
" data_router=engine.router,\n",
")\n",
2026-03-09 22:33:41 +08:00
"print(\"[股票池筛选] 使用自定义函数进行股票池筛选\")\n",
"print(f\"[股票池筛选] 所需基础列: {STOCK_FILTER_REQUIRED_COLUMNS}\")\n",
"print(\"[股票池筛选] 筛选逻辑: 排除创业板/科创板/北交所后，
"# print(f\"[股票池筛选] 所需因子: {list(STOCK_FILTER_REQUIRED_FACTORS.keys())}\")\n",
2026-03-06 20:57:27 +08:00
"\n",
"# 9. 创建 ST 股票过滤器\n",
"st_filter = STFilter(\n",
" data_router=engine.router,\n",
")\n",
"\n",
"# 10. 创建训练器\n",
"trainer = Trainer(\n",
" model=model,\n",
" pool_manager=pool_manager,\n",
" processors=processors,\n",
2026-03-09 22:33:41 +08:00
" filters=[st_filter], # 使用STFilter过滤ST股票\n",
2026-03-06 20:57:27 +08:00
" splitter=splitter,\n",
" target_col=target_col,\n",
" feature_cols=feature_cols,\n",
" persist_model=PERSIST_MODEL,\n",
")"
],
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"================================================================================\n",
"LightGBM 回归模型训练\n",
"================================================================================\n",
"\n",
"[1] 创建 FactorEngine\n",
"\n",
"[2] 定义因子（字符串表达式）\n",
"================================================================================\n",
"使用字符串表达式定义因子\n",
"================================================================================\n",
"\n",
"注册特征因子:\n",
2026-03-08 23:14:18 +08:00
" - ma_5: ts_mean(close, 5)\n",
" - ma_20: ts_mean(close, 20)\n",
" - ma_ratio_5_20: ts_mean(close, 5) / (ts_mean(close, 20) + 1e-8) - 1\n",
" - bias_10: close / (ts_mean(close, 10) + 1e-8) - 1\n",
2026-03-09 23:37:20 +08:00
" - high_low_ratio: (close - ts_min(low, 20)) / (ts_max(high, 20) - ts_min(low, 20) + 1e-8)\n",
2026-03-08 23:14:18 +08:00
" - bbi_ratio: (ts_mean(close, 3) + ts_mean(close, 6) + ts_mean(close, 12) + ts_mean(close, 24)) / (4 * close + 1e-8)\n",
2026-03-09 23:37:20 +08:00
" - return_5: (close / (ts_delay(close, 5) + 1e-8)) - 1\n",
" - return_20: (close / (ts_delay(close, 20) + 1e-8)) - 1\n",
" - kaufman_ER_20: abs(close - ts_delay(close, 20)) / (ts_sum(abs(close - ts_delay(close, 1)), 20) + 1e-8)\n",
" - mom_acceleration_10_20: (close / (ts_delay(close, 10) + 1e-8) - 1) - (ts_delay(close, 10) / (ts_delay(close, 20) + 1e-8) - 1)\n",
" - drawdown_from_high_60: close / (ts_max(high, 60) + 1e-8) - 1\n",
" - up_days_ratio_20: ts_sum(close > ts_delay(close, 1), 20) / 20\n",
" - volatility_5: ts_std(close, 5)\n",
" - volatility_20: ts_std(close, 20)\n",
2026-03-08 23:14:18 +08:00
" - volatility_ratio: ts_std(close, 5) / (ts_std(close, 20) + 1e-8)\n",
2026-03-09 23:37:20 +08:00
" - std_return_20: ts_std((close / (ts_delay(close, 1) + 1e-8)) - 1, 20)\n",
" - sharpe_ratio_20: ts_mean(close / (ts_delay(close, 1) + 1e-8) - 1, 20) / (ts_std(close / (ts_delay(close, 1) + 1e-8) - 1, 20) + 1e-8)\n",
" - min_ret_20: ts_min(close / (ts_delay(close, 1) + 1e-8) - 1, 20)\n",
" - volatility_squeeze_5_60: ts_std(close, 5) / (ts_std(close, 60) + 1e-8)\n",
" - overnight_intraday_diff: (open / (ts_delay(close, 1) + 1e-8) - 1) - (close / (open + 1e-8) - 1)\n",
" - upper_shadow_ratio: (high - ((open + close + abs(open - close)) / 2)) / (high - low + 1e-8)\n",
" - capital_retention_20: ts_sum(abs(close - open), 20) / (ts_sum(high - low, 20) + 1e-8)\n",
" - max_ret_20: ts_max(close / (ts_delay(close, 1) + 1e-8) - 1, 20)\n",
" - volume_ratio_5_20: ts_mean(vol, 5) / (ts_mean(vol, 20) + 1e-8)\n",
2026-03-08 14:12:03 +08:00
" - turnover_rate_mean_5: ts_mean(turnover_rate, 5)\n",
2026-03-08 23:14:18 +08:00
" - turnover_deviation: (turnover_rate - ts_mean(turnover_rate, 10)) / (ts_std(turnover_rate, 10) + 1e-8)\n",
2026-03-09 23:37:20 +08:00
" - amihud_illiq_20: ts_mean(abs(close / (ts_delay(close, 1) + 1e-8) - 1) / (amount + 1e-8), 20)\n",
" - turnover_cv_20: ts_std(turnover_rate, 20) / (ts_mean(turnover_rate, 20) + 1e-8)\n",
" - pv_corr_20: ts_corr(close / (ts_delay(close, 1) + 1e-8) - 1, vol, 20)\n",
" - close_vwap_deviation: close / (amount / (vol * 100 + 1e-8) + 1e-8) - 1\n",
2026-03-09 22:33:41 +08:00
" - roe: n_income / (total_hldr_eqy_exc_min_int + 1e-8)\n",
2026-03-09 23:37:20 +08:00
" - roa: n_income / (total_assets + 1e-8)\n",
" - profit_margin: n_income / (revenue + 1e-8)\n",
2026-03-09 22:33:41 +08:00
" - debt_to_equity: total_liab / (total_hldr_eqy_exc_min_int + 1e-8)\n",
" - current_ratio: total_cur_assets / (total_cur_liab + 1e-8)\n",
2026-03-09 23:37:20 +08:00
" - net_profit_yoy: (n_income / (ts_delay(n_income, 252) + 1e-8)) - 1\n",
" - revenue_yoy: (revenue / (ts_delay(revenue, 252) + 1e-8)) - 1\n",
" - healthy_expansion_velocity: (total_assets / (ts_delay(total_assets, 252) + 1e-8) - 1) - (total_liab / (ts_delay(total_liab, 252) + 1e-8) - 1)\n",
" - EP: n_income / (total_mv * 10000 + 1e-8)\n",
" - BP: total_hldr_eqy_exc_min_int / (total_mv * 10000 + 1e-8)\n",
" - CP: n_cashflow_act / (total_mv * 10000 + 1e-8)\n",
2026-03-08 23:14:18 +08:00
" - market_cap_rank: cs_rank(total_mv)\n",
2026-03-09 23:37:20 +08:00
" - turnover_rank: cs_rank(turnover_rate)\n",
" - return_5_rank: cs_rank((close / (ts_delay(close, 5) + 1e-8)) - 1)\n",
" - EP_rank: cs_rank(n_income / (total_mv + 1e-8))\n",
" - pe_expansion_trend: (total_mv / (n_income + 1e-8)) / (ts_delay(total_mv, 60) / (ts_delay(n_income, 60) + 1e-8) + 1e-8) - 1\n",
" - value_price_divergence: cs_rank((n_income - ts_delay(n_income, 252)) / (abs(ts_delay(n_income, 252)) + 1e-8)) - cs_rank(close / (ts_delay(close, 20) + 1e-8))\n",
" - active_market_cap: total_mv * ts_mean(turnover_rate, 20)\n",
" - ebit_rank: cs_rank(ebit)\n",
2026-03-06 20:57:27 +08:00
"\n",
"注册 Label 因子:\n",
2026-03-09 22:33:41 +08:00
" - future_return_5: (ts_delay(close, -5) / ts_delay(open, -1)) - 1\n",
2026-03-06 20:57:27 +08:00
"\n",
2026-03-09 23:37:20 +08:00
"特征因子数: 49\n",
2026-03-08 14:12:03 +08:00
"Label: future_return_5\n",
2026-03-09 23:37:20 +08:00
"已注册因子总数: 50\n",
2026-03-06 20:57:27 +08:00
"\n",
"[3] 准备数据\n",
"\n",
"================================================================================\n",
"准备数据\n",
"================================================================================\n",
"\n",
2026-03-09 22:33:41 +08:00
"计算因子: 20200101 - 20261231\n",
"[FinancialLoader] 加载 financial_fina_indicator 失败: Binder Error: Referenced column \"f_ann_date\" not found in FROM clause!\n",
"Candidate bindings: \"ann_date\", \"end_date\", \"ocf_to_debt\", \"arturn_days\", \"nca_to_assets\"\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"D:\\PyProject\\ProStock\\src\\data\\financial_loader.py:148: UserWarning: Sortedness of columns cannot be checked when 'by' groups provided\n",
" merged = df_price.join_asof(\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
2026-03-09 23:37:20 +08:00
"数据形状: (7255513, 70)\n",
"数据列: ['ts_code', 'trade_date', 'high', 'low', 'amount', 'vol', 'close', 'turnover_rate', 'open', 'total_assets', 'total_mv', 'f_ann_date', 'total_cur_liab', 'total_cur_assets', 'total_hldr_eqy_exc_min_int', 'total_liab', 'revenue', 'n_income', 'n_cashflow_act', 'ebit', 'ma_5', 'ma_20', 'ma_ratio_5_20', 'bias_10', 'high_low_ratio', 'bbi_ratio', 'return_5', 'return_20', 'kaufman_ER_20', 'mom_acceleration_10_20', 'drawdown_from_high_60', 'up_days_ratio_20', 'volatility_5', 'volatility_20', 'volatility_ratio', 'std_return_20', 'sharpe_ratio_20', 'min_ret_20', 'volatility_squeeze_5_60', 'overnight_intraday_diff', 'upper_shadow_ratio', 'capital_retention_20', 'max_ret_20', 'volume_ratio_5_20', 'turnover_rate_mean_5', 'turnover_deviation', 'amihud_illiq_20', 'turnover_cv_20', 'pv_corr_20', 'close_vwap_deviation', 'roe', 'roa', 'profit_margin', 'debt_to_equity', 'current_ratio', 'net_profit_yoy', 'revenue_yoy', 'healthy_expansion_velocity', 'EP', 'BP', 'CP', 'market_cap_rank', 'turnover_rank', 'return_5_rank', 'EP_rank', 'pe_expansion_trend', 'value_price_divergence', 'active_market_cap', 'ebit_rank', 'future_return_5']\n",
2026-03-06 20:57:27 +08:00
"\n",
"前5行预览:\n",
2026-03-09 23:37:20 +08:00
"shape: (5, 70)\n",
"┌───────────┬────────────┬─────────┬─────────┬───┬────────────┬────────────┬───────────┬───────────┐\n",
"│ ts_code ┆ trade_date ┆ high ┆ low ┆ … ┆ value_pric ┆ active_mar ┆ ebit_rank ┆ future_re │\n",
"│ --- ┆ --- ┆ --- ┆ --- ┆ ┆ e_divergen ┆ ket_cap ┆ --- ┆ turn_5 │\n",
"│ str ┆ str ┆ f64 ┆ f64 ┆ ┆ ce ┆ --- ┆ f64 ┆ --- │\n",
"│ ┆ ┆ ┆ ┆ ┆ --- ┆ f64 ┆ ┆ f64 │\n",
"│ ┆ ┆ ┆ ┆ ┆ f64 ┆ ┆ ┆ │\n",
"╞═══════════╪════════════╪═════════╪═════════╪═══╪════════════╪════════════╪═══════════╪═══════════╡\n",
"│ 000001.SZ ┆ 20200102 ┆ 1850.42 ┆ 1806.75 ┆ … ┆ null ┆ null ┆ null ┆ -0.008857 │\n",
"│ 000001.SZ ┆ 20200103 ┆ 1889.72 ┆ 1847.15 ┆ … ┆ null ┆ null ┆ null ┆ -0.01881 │\n",
"│ 000001.SZ ┆ 20200106 ┆ 1893.0 ┆ 1846.05 ┆ … ┆ null ┆ null ┆ null ┆ -0.008171 │\n",
"│ 000001.SZ ┆ 20200107 ┆ 1886.45 ┆ 1850.42 ┆ … ┆ null ┆ null ┆ null ┆ -0.014117 │\n",
"│ 000001.SZ ┆ 20200108 ┆ 1861.34 ┆ 1815.49 ┆ … ┆ null ┆ null ┆ null ┆ -0.017252 │\n",
"└───────────┴────────────┴─────────┴─────────┴───┴────────────┴────────────┴───────────┴───────────┘\n",
2026-03-06 20:57:27 +08:00
"\n",
2026-03-08 01:09:47 +08:00
"[配置] 训练期: 20200101 - 20231231\n",
"[配置] 验证期: 20240101 - 20241231\n",
2026-03-09 22:33:41 +08:00
"[配置] 测试期: 20250101 - 20261231\n",
2026-03-09 23:37:20 +08:00
"[配置] 特征数: 49\n",
2026-03-09 22:33:41 +08:00
"[配置] 目标变量: future_return_5\n",
"[股票池筛选] 使用自定义函数进行股票池筛选\n",
"[股票池筛选] 所需基础列: ['total_mv']\n",
"[股票池筛选] 筛选逻辑: 排除创业板/科创板/北交所后，
2026-03-06 20:57:27 +08:00
]
}
],
2026-03-09 22:33:41 +08:00
"execution_count": 5
2026-03-06 20:57:27 +08:00
},
{
"metadata": {},
2026-03-09 22:33:41 +08:00
"cell_type": "markdown",
"source": "### 4.2 执行训练"
2026-03-06 20:57:27 +08:00
},
{
"metadata": {
"ExecuteTime": {
2026-03-09 23:37:20 +08:00
"end_time": "2026-03-09T15:27:01.856502Z",
"start_time": "2026-03-09T15:26:58.307450Z"
2026-03-06 20:57:27 +08:00
}
},
2026-03-09 22:33:41 +08:00
"cell_type": "code",
2026-03-06 20:57:27 +08:00
"source": [
"print(\"\\n\" + \"=\" * 80)\n",
"print(\"开始训练\")\n",
"print(\"=\" * 80)\n",
"\n",
"# 步骤 1: 股票池筛选\n",
"print(\"\\n[步骤 1/6] 股票池筛选\")\n",
"print(\"-\" * 60)\n",
"if pool_manager:\n",
" print(\" 执行每日独立筛选股票池...\")\n",
" filtered_data = pool_manager.filter_and_select_daily(data)\n",
" print(f\" 筛选前数据规模: {data.shape}\")\n",
" print(f\" 筛选后数据规模: {filtered_data.shape}\")\n",
" print(f\" 筛选前股票数: {data['ts_code'].n_unique()}\")\n",
" print(f\" 筛选后股票数: {filtered_data['ts_code'].n_unique()}\")\n",
" print(f\" 删除记录数: {len(data) - len(filtered_data)}\")\n",
"else:\n",
" filtered_data = data\n",
" print(\" 未配置股票池管理器，跳过筛选\")"
],
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"================================================================================\n",
"开始训练\n",
"================================================================================\n",
"\n",
"[步骤 1/6] 股票池筛选\n",
"------------------------------------------------------------\n",
2026-03-09 22:33:41 +08:00
" 执行每日独立筛选股票池...\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
2026-03-09 23:37:20 +08:00
"C:\\Users\\liaozhaorun\\AppData\\Local\\Temp\\ipykernel_37796\\547547317.py:75: DeprecationWarning: `is_in` with a collection of the same datatype is ambiguous and deprecated.\n",
2026-03-09 22:33:41 +08:00
"Please use `implode` to return to previous behavior.\n",
"\n",
"See https://github.com/pola-rs/polars/issues/22149 for more information.\n",
" return df[\"ts_code\"].is_in(small_cap_codes)\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
2026-03-09 23:37:20 +08:00
" 筛选前数据规模: (7255513, 70)\n",
" 筛选后数据规模: (1494000, 70)\n",
2026-03-09 22:33:41 +08:00
" 筛选前股票数: 5694\n",
" 筛选后股票数: 2252\n",
" 删除记录数: 5761513\n"
2026-03-06 20:57:27 +08:00
]
}
],
2026-03-09 22:33:41 +08:00
"execution_count": 6
2026-03-06 20:57:27 +08:00
},
{
"metadata": {
"ExecuteTime": {
2026-03-09 23:37:20 +08:00
"end_time": "2026-03-09T15:27:01.935567Z",
"start_time": "2026-03-09T15:27:01.869943Z"
2026-03-06 20:57:27 +08:00
}
},
2026-03-09 22:33:41 +08:00
"cell_type": "code",
2026-03-06 20:57:27 +08:00
"source": [
2026-03-08 01:09:47 +08:00
"# 步骤 2: 划分训练/验证/测试集（正确的三分法）\n",
"print(\"\\n[步骤 2/6] 划分训练集、验证集和测试集\")\n",
2026-03-06 20:57:27 +08:00
"print(\"-\" * 60)\n",
"if splitter:\n",
2026-03-08 01:09:47 +08:00
" # 正确的三分法： ， ，
" train_data, val_data, test_data = splitter.split(filtered_data)\n",
2026-03-06 20:57:27 +08:00
" print(f\" 训练集数据规模: {train_data.shape}\")\n",
2026-03-08 01:09:47 +08:00
" print(f\" 验证集数据规模: {val_data.shape}\")\n",
2026-03-06 20:57:27 +08:00
" print(f\" 测试集数据规模: {test_data.shape}\")\n",
" print(f\" 训练集股票数: {train_data['ts_code'].n_unique()}\")\n",
2026-03-08 01:09:47 +08:00
" print(f\" 验证集股票数: {val_data['ts_code'].n_unique()}\")\n",
2026-03-06 20:57:27 +08:00
" print(f\" 测试集股票数: {test_data['ts_code'].n_unique()}\")\n",
" print(\n",
" f\" 训练集日期范围: {train_data['trade_date'].min()} - {train_data['trade_date'].max()}\"\n",
" )\n",
" print(\n",
2026-03-08 01:09:47 +08:00
" f\" 验证集日期范围: {val_data['trade_date'].min()} - {val_data['trade_date'].max()}\"\n",
" )\n",
" print(\n",
2026-03-06 20:57:27 +08:00
" f\" 测试集日期范围: {test_data['trade_date'].min()} - {test_data['trade_date'].max()}\"\n",
" )\n",
"\n",
" print(\"\\n 训练集前5行预览:\")\n",
" print(train_data.head())\n",
2026-03-08 01:09:47 +08:00
" print(\"\\n 验证集前5行预览:\")\n",
" print(val_data.head())\n",
2026-03-06 20:57:27 +08:00
" print(\"\\n 测试集前5行预览:\")\n",
" print(test_data.head())\n",
"else:\n",
" train_data = filtered_data\n",
" test_data = filtered_data\n",
" print(\" 未配置划分器，全部作为训练集\")"
],
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
2026-03-08 01:09:47 +08:00
"[步骤 2/6] 划分训练集、验证集和测试集\n",
2026-03-06 20:57:27 +08:00
"------------------------------------------------------------\n",
2026-03-09 23:37:20 +08:00
" 训练集数据规模: (970000, 70)\n",
" 验证集数据规模: (242000, 70)\n",
" 测试集数据规模: (282000, 70)\n",
2026-03-09 22:33:41 +08:00
" 训练集股票数: 1888\n",
" 验证集股票数: 1377\n",
" 测试集股票数: 1682\n",
2026-03-08 01:09:47 +08:00
" 训练集日期范围: 20200102 - 20231229\n",
" 验证集日期范围: 20240102 - 20241231\n",
2026-03-09 22:33:41 +08:00
" 测试集日期范围: 20250102 - 20260306\n",
2026-03-06 20:57:27 +08:00
"\n",
" 训练集前5行预览:\n",
2026-03-09 23:37:20 +08:00
"shape: (5, 70)\n",
"┌───────────┬────────────┬───────┬───────┬───┬─────────────┬─────────────┬───────────┬─────────────┐\n",
"│ ts_code ┆ trade_date ┆ high ┆ low ┆ … ┆ value_price ┆ active_mark ┆ ebit_rank ┆ future_retu │\n",
"│ --- ┆ --- ┆ --- ┆ --- ┆ ┆ _divergence ┆ et_cap ┆ --- ┆ rn_5 │\n",
"│ str ┆ str ┆ f64 ┆ f64 ┆ ┆ --- ┆ --- ┆ f64 ┆ --- │\n",
"│ ┆ ┆ ┆ ┆ ┆ f64 ┆ f64 ┆ ┆ f64 │\n",
"╞═══════════╪════════════╪═══════╪═══════╪═══╪═════════════╪═════════════╪═══════════╪═════════════╡\n",
"│ 000004.SZ ┆ 20200102 ┆ 93.06 ┆ 90.1 ┆ … ┆ null ┆ null ┆ null ┆ 0.000441 │\n",
"│ 000004.SZ ┆ 20200103 ┆ 91.36 ┆ 89.53 ┆ … ┆ null ┆ null ┆ null ┆ 0.005875 │\n",
"│ 000004.SZ ┆ 20200106 ┆ 90.22 ┆ 87.58 ┆ … ┆ null ┆ null ┆ null ┆ 0.05644 │\n",
"│ 000004.SZ ┆ 20200107 ┆ 90.22 ┆ 88.06 ┆ … ┆ null ┆ null ┆ null ┆ 0.049753 │\n",
"│ 000004.SZ ┆ 20200108 ┆ 92.54 ┆ 88.51 ┆ … ┆ null ┆ null ┆ null ┆ 0.019922 │\n",
"└───────────┴────────────┴───────┴───────┴───┴─────────────┴─────────────┴───────────┴─────────────┘\n",
2026-03-08 01:09:47 +08:00
"\n",
" 验证集前5行预览:\n",
2026-03-09 23:37:20 +08:00
"shape: (5, 70)\n",
"┌───────────┬────────────┬───────┬───────┬───┬─────────────┬─────────────┬───────────┬─────────────┐\n",
"│ ts_code ┆ trade_date ┆ high ┆ low ┆ … ┆ value_price ┆ active_mark ┆ ebit_rank ┆ future_retu │\n",
"│ --- ┆ --- ┆ --- ┆ --- ┆ ┆ _divergence ┆ et_cap ┆ --- ┆ rn_5 │\n",
"│ str ┆ str ┆ f64 ┆ f64 ┆ ┆ --- ┆ --- ┆ f64 ┆ --- │\n",
"│ ┆ ┆ ┆ ┆ ┆ f64 ┆ f64 ┆ ┆ f64 │\n",
"╞═══════════╪════════════╪═══════╪═══════╪═══╪═════════════╪═════════════╪═══════════╪═════════════╡\n",
"│ 000004.SZ ┆ 20240102 ┆ 66.81 ┆ 65.23 ┆ … ┆ null ┆ 770442.9948 ┆ null ┆ -0.014188 │\n",
"│ ┆ ┆ ┆ ┆ ┆ ┆ 33 ┆ ┆ │\n",
"│ 000004.SZ ┆ 20240103 ┆ 66.24 ┆ 64.62 ┆ … ┆ null ┆ 751492.2017 ┆ null ┆ 0.002432 │\n",
"│ ┆ ┆ ┆ ┆ ┆ ┆ 8 ┆ ┆ │\n",
"│ 000004.SZ ┆ 20240104 ┆ 71.24 ┆ 64.7 ┆ … ┆ null ┆ 866443.5445 ┆ null ┆ 0.016919 │\n",
"│ ┆ ┆ ┆ ┆ ┆ ┆ 25 ┆ ┆ │\n",
"│ 000004.SZ ┆ 20240105 ┆ 71.08 ┆ 65.19 ┆ … ┆ null ┆ 907980.5905 ┆ null ┆ -0.013477 │\n",
"│ ┆ ┆ ┆ ┆ ┆ ┆ 95 ┆ ┆ │\n",
"│ 000004.SZ ┆ 20240108 ┆ 67.87 ┆ 65.02 ┆ … ┆ null ┆ 931205.3950 ┆ null ┆ -0.024684 │\n",
"│ ┆ ┆ ┆ ┆ ┆ ┆ 63 ┆ ┆ │\n",
"└───────────┴────────────┴───────┴───────┴───┴─────────────┴─────────────┴───────────┴─────────────┘\n",
2026-03-06 20:57:27 +08:00
"\n",
" 测试集前5行预览:\n",
2026-03-09 23:37:20 +08:00
"shape: (5, 70)\n",
"┌───────────┬────────────┬───────┬───────┬───┬─────────────┬─────────────┬───────────┬─────────────┐\n",
"│ ts_code ┆ trade_date ┆ high ┆ low ┆ … ┆ value_price ┆ active_mark ┆ ebit_rank ┆ future_retu │\n",
"│ --- ┆ --- ┆ --- ┆ --- ┆ ┆ _divergence ┆ et_cap ┆ --- ┆ rn_5 │\n",
"│ str ┆ str ┆ f64 ┆ f64 ┆ ┆ --- ┆ --- ┆ f64 ┆ --- │\n",
"│ ┆ ┆ ┆ ┆ ┆ f64 ┆ f64 ┆ ┆ f64 │\n",
"╞═══════════╪════════════╪═══════╪═══════╪═══╪═════════════╪═════════════╪═══════════╪═════════════╡\n",
"│ 000004.SZ ┆ 20250102 ┆ 58.48 ┆ 54.17 ┆ … ┆ null ┆ 2.3754e6 ┆ null ┆ -0.066193 │\n",
"│ 000004.SZ ┆ 20250103 ┆ 58.52 ┆ 51.86 ┆ … ┆ null ┆ 2.1884e6 ┆ null ┆ 0.00893 │\n",
"│ 000004.SZ ┆ 20250106 ┆ 52.55 ┆ 49.17 ┆ … ┆ null ┆ 2.1549e6 ┆ null ┆ -0.0142 │\n",
"│ 000004.SZ ┆ 20250107 ┆ 53.32 ┆ 51.41 ┆ … ┆ null ┆ 2.2770e6 ┆ null ┆ 0.013031 │\n",
"│ 000004.SZ ┆ 20250108 ┆ 54.78 ┆ 52.38 ┆ … ┆ null ┆ 2.3533e6 ┆ null ┆ 0.00442 │\n",
"└───────────┴────────────┴───────┴───────┴───┴─────────────┴─────────────┴───────────┴─────────────┘\n"
2026-03-06 20:57:27 +08:00
]
}
],
2026-03-09 22:33:41 +08:00
"execution_count": 7
2026-03-06 20:57:27 +08:00
},
{
"metadata": {
"ExecuteTime": {
2026-03-09 23:37:20 +08:00
"end_time": "2026-03-09T15:27:02.522387Z",
"start_time": "2026-03-09T15:27:01.942140Z"
2026-03-06 20:57:27 +08:00
}
},
2026-03-09 22:33:41 +08:00
"cell_type": "code",
2026-03-06 20:57:27 +08:00
"source": [
"# 步骤 3: 训练集数据处理\n",
"print(\"\\n[步骤 3/6] 训练集数据处理\")\n",
"print(\"-\" * 60)\n",
"fitted_processors = []\n",
"if processors:\n",
" for i, processor in enumerate(processors, 1):\n",
" print(\n",
" f\" [{i}/{len(processors)}] 应用处理器: {processor.__class__.__name__}\"\n",
" )\n",
" train_data_before = len(train_data)\n",
" train_data = processor.fit_transform(train_data)\n",
" train_data_after = len(train_data)\n",
" fitted_processors.append(processor)\n",
" print(f\" 处理前记录数: {train_data_before}\")\n",
" print(f\" 处理后记录数: {train_data_after}\")\n",
" if train_data_before != train_data_after:\n",
" print(f\" 删除记录数: {train_data_before - train_data_after}\")\n",
"\n",
"print(\"\\n 训练集处理后前5行预览:\")\n",
"print(train_data.head())\n",
"print(f\"\\n 训练集特征统计:\")\n",
"print(f\" 特征数: {len(feature_cols)}\")\n",
"print(f\" 样本数: {len(train_data)}\")\n",
"print(f\" 缺失值统计:\")\n",
"for col in feature_cols[:5]: # 只显示前5个特征的缺失值\n",
" null_count = train_data[col].null_count()\n",
" if null_count > 0:\n",
" print(\n",
" f\" {col}: {null_count} ({null_count / len(train_data) * 100:.2f}%)\"\n",
" )"
],
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"[步骤 3/6] 训练集数据处理\n",
"------------------------------------------------------------\n",
" [1/3] 应用处理器: NullFiller\n",
2026-03-09 22:33:41 +08:00
" 处理前记录数: 970000\n",
" 处理后记录数: 970000\n",
2026-03-06 20:57:27 +08:00
" [2/3] 应用处理器: Winsorizer\n",
2026-03-09 22:33:41 +08:00
" 处理前记录数: 970000\n",
" 处理后记录数: 970000\n",
2026-03-06 20:57:27 +08:00
" [3/3] 应用处理器: StandardScaler\n",
2026-03-09 22:33:41 +08:00
" 处理前记录数: 970000\n",
" 处理后记录数: 970000\n",
2026-03-06 20:57:27 +08:00
"\n",
" 训练集处理后前5行预览:\n",
2026-03-09 23:37:20 +08:00
"shape: (5, 70)\n",
"┌───────────┬────────────┬──────────┬──────────┬───┬───────────┬───────────┬───────────┬───────────┐\n",
"│ ts_code ┆ trade_date ┆ high ┆ low ┆ … ┆ value_pri ┆ active_ma ┆ ebit_rank ┆ future_re │\n",
"│ --- ┆ --- ┆ --- ┆ --- ┆ ┆ ce_diverg ┆ rket_cap ┆ --- ┆ turn_5 │\n",
"│ str ┆ str ┆ f64 ┆ f64 ┆ ┆ ence ┆ --- ┆ f64 ┆ --- │\n",
"│ ┆ ┆ ┆ ┆ ┆ --- ┆ f64 ┆ ┆ f64 │\n",
"│ ┆ ┆ ┆ ┆ ┆ f64 ┆ ┆ ┆ │\n",
"╞═══════════╪════════════╪══════════╪══════════╪═══╪═══════════╪═══════════╪═══════════╪═══════════╡\n",
"│ 000004.SZ ┆ 20200102 ┆ 4.202576 ┆ 4.206515 ┆ … ┆ null ┆ null ┆ null ┆ 0.000441 │\n",
"│ 000004.SZ ┆ 20200103 ┆ 4.202576 ┆ 4.206515 ┆ … ┆ null ┆ null ┆ null ┆ 0.005875 │\n",
"│ 000004.SZ ┆ 20200106 ┆ 4.202576 ┆ 4.206515 ┆ … ┆ null ┆ null ┆ null ┆ 0.05644 │\n",
"│ 000004.SZ ┆ 20200107 ┆ 4.202576 ┆ 4.206515 ┆ … ┆ null ┆ null ┆ null ┆ 0.049753 │\n",
"│ 000004.SZ ┆ 20200108 ┆ 4.202576 ┆ 4.206515 ┆ … ┆ null ┆ null ┆ null ┆ 0.019922 │\n",
"└───────────┴────────────┴──────────┴──────────┴───┴───────────┴───────────┴───────────┴───────────┘\n",
2026-03-06 20:57:27 +08:00
"\n",
" 训练集特征统计:\n",
2026-03-09 23:37:20 +08:00
" 特征数: 49\n",
2026-03-09 22:33:41 +08:00
" 样本数: 970000\n",
2026-03-06 20:57:27 +08:00
" 缺失值统计:\n",
2026-03-09 22:33:41 +08:00
" ma_5: 4000 (0.41%)\n",
" ma_20: 19000 (1.96%)\n",
" ma_ratio_5_20: 19000 (1.96%)\n",
" bias_10: 9000 (0.93%)\n",
2026-03-09 23:37:20 +08:00
" high_low_ratio: 19000 (1.96%)\n"
2026-03-06 20:57:27 +08:00
]
}
],
2026-03-09 22:33:41 +08:00
"execution_count": 8
2026-03-06 20:57:27 +08:00
},
{
"metadata": {
"ExecuteTime": {
2026-03-09 23:37:20 +08:00
"end_time": "2026-03-09T15:27:10.946482Z",
"start_time": "2026-03-09T15:27:02.528042Z"
2026-03-06 20:57:27 +08:00
}
},
2026-03-09 22:33:41 +08:00
"cell_type": "code",
2026-03-06 20:57:27 +08:00
"source": [
"# 步骤 4: 训练模型\n",
"print(\"\\n[步骤 4/6] 训练模型\")\n",
"print(\"-\" * 60)\n",
"print(f\" 模型类型: LightGBM\")\n",
"print(f\" 训练样本数: {len(train_data)}\")\n",
"print(f\" 特征数: {len(feature_cols)}\")\n",
"print(f\" 目标变量: {target_col}\")\n",
"\n",
"X_train = train_data.select(feature_cols)\n",
"y_train = train_data.select(target_col).to_series()\n",
"\n",
"print(f\"\\n 目标变量统计:\")\n",
"print(f\" 均值: {y_train.mean():.6f}\")\n",
"print(f\" 标准差: {y_train.std():.6f}\")\n",
"print(f\" 最小值: {y_train.min():.6f}\")\n",
"print(f\" 最大值: {y_train.max():.6f}\")\n",
"print(f\" 缺失值: {y_train.null_count()}\")\n",
"\n",
"print(\"\\n 开始训练...\")\n",
"model.fit(X_train, y_train)\n",
"print(\" 训练完成！\")"
],
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"[步骤 4/6] 训练模型\n",
"------------------------------------------------------------\n",
" 模型类型: LightGBM\n",
2026-03-09 22:33:41 +08:00
" 训练样本数: 970000\n",
2026-03-09 23:37:20 +08:00
" 特征数: 49\n",
2026-03-08 14:12:03 +08:00
" 目标变量: future_return_5\n",
2026-03-06 20:57:27 +08:00
"\n",
" 目标变量统计:\n",
2026-03-09 22:33:41 +08:00
" 均值: 0.004184\n",
" 标准差: 0.058740\n",
" 最小值: -0.152621\n",
" 最大值: 0.216472\n",
2026-03-06 20:57:27 +08:00
" 缺失值: 0\n",
"\n",
" 开始训练...\n",
" 训练完成！\n"
]
}
],
2026-03-09 22:33:41 +08:00
"execution_count": 9
2026-03-06 20:57:27 +08:00
},
{
"metadata": {
"ExecuteTime": {
2026-03-09 23:37:20 +08:00
"end_time": "2026-03-09T15:27:10.997637Z",
"start_time": "2026-03-09T15:27:10.951980Z"
2026-03-06 20:57:27 +08:00
}
},
2026-03-09 22:33:41 +08:00
"cell_type": "code",
2026-03-06 20:57:27 +08:00
"source": [
"# 步骤 5: 测试集数据处理\n",
"print(\"\\n[步骤 5/6] 测试集数据处理\")\n",
"print(\"-\" * 60)\n",
"if processors and test_data is not train_data:\n",
" for i, processor in enumerate(fitted_processors, 1):\n",
" print(\n",
" f\" [{i}/{len(fitted_processors)}] 应用处理器: {processor.__class__.__name__}\"\n",
" )\n",
" test_data_before = len(test_data)\n",
" test_data = processor.transform(test_data)\n",
" test_data_after = len(test_data)\n",
" print(f\" 处理前记录数: {test_data_before}\")\n",
" print(f\" 处理后记录数: {test_data_after}\")\n",
"else:\n",
" print(\" 跳过测试集处理\")"
],
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"[步骤 5/6] 测试集数据处理\n",
"------------------------------------------------------------\n",
" [1/3] 应用处理器: NullFiller\n",
2026-03-09 22:33:41 +08:00
" 处理前记录数: 282000\n",
" 处理后记录数: 282000\n",
2026-03-06 20:57:27 +08:00
" [2/3] 应用处理器: Winsorizer\n",
2026-03-09 22:33:41 +08:00
" 处理前记录数: 282000\n",
" 处理后记录数: 282000\n",
2026-03-06 20:57:27 +08:00
" [3/3] 应用处理器: StandardScaler\n",
2026-03-09 22:33:41 +08:00
" 处理前记录数: 282000\n",
" 处理后记录数: 282000\n"
2026-03-06 20:57:27 +08:00
]
}
],
2026-03-09 22:33:41 +08:00
"execution_count": 10
2026-03-06 20:57:27 +08:00
},
{
"metadata": {
"ExecuteTime": {
2026-03-09 23:37:20 +08:00
"end_time": "2026-03-09T15:27:11.481411Z",
"start_time": "2026-03-09T15:27:11.004438Z"
2026-03-06 20:57:27 +08:00
}
},
2026-03-09 22:33:41 +08:00
"cell_type": "code",
2026-03-06 20:57:27 +08:00
"source": [
"# 步骤 6: 生成预测\n",
"print(\"\\n[步骤 6/6] 生成预测\")\n",
"print(\"-\" * 60)\n",
"X_test = test_data.select(feature_cols)\n",
"print(f\" 测试样本数: {len(X_test)}\")\n",
"print(\" 预测中...\")\n",
"predictions = model.predict(X_test)\n",
"print(f\" 预测完成！\")\n",
"\n",
"print(f\"\\n 预测结果统计:\")\n",
"print(f\" 均值: {predictions.mean():.6f}\")\n",
"print(f\" 标准差: {predictions.std():.6f}\")\n",
"print(f\" 最小值: {predictions.min():.6f}\")\n",
"print(f\" 最大值: {predictions.max():.6f}\")\n",
"\n",
"# 保存结果到 trainer\n",
"trainer.results = test_data.with_columns([pl.Series(\"prediction\", predictions)])"
],
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"[步骤 6/6] 生成预测\n",
"------------------------------------------------------------\n",
2026-03-09 22:33:41 +08:00
" 测试样本数: 282000\n",
2026-03-06 20:57:27 +08:00
" 预测中...\n",
" 预测完成！\n",
"\n",
" 预测结果统计:\n",
2026-03-09 23:37:20 +08:00
" 均值: 0.002266\n",
" 标准差: 0.009319\n",
" 最小值: -0.117614\n",
" 最大值: 0.080849\n"
2026-03-08 01:09:47 +08:00
]
}
],
2026-03-09 22:33:41 +08:00
"execution_count": 11
2026-03-08 01:09:47 +08:00
},
{
"metadata": {},
2026-03-09 22:33:41 +08:00
"cell_type": "markdown",
"source": "### 4.3 训练指标曲线"
2026-03-08 01:09:47 +08:00
},
{
"metadata": {
"ExecuteTime": {
2026-03-09 23:37:20 +08:00
"end_time": "2026-03-09T15:27:15.029980Z",
"start_time": "2026-03-09T15:27:11.487153Z"
2026-03-08 01:09:47 +08:00
}
},
2026-03-09 22:33:41 +08:00
"cell_type": "code",
2026-03-08 01:09:47 +08:00
"source": [
"print(\"\\n\" + \"=\" * 80)\n",
"print(\"训练指标曲线\")\n",
"print(\"=\" * 80)\n",
"\n",
"# 重新训练以收集指标（因为之前的训练没有保存评估结果）\n",
"print(\"\\n重新训练模型以收集训练指标...\")\n",
"\n",
"import lightgbm as lgb\n",
"\n",
"# 准备数据（使用 val 做验证，
"X_train_np = X_train.to_numpy()\n",
"y_train_np = y_train.to_numpy()\n",
"X_val_np = val_data.select(feature_cols).to_numpy()\n",
"y_val_np = val_data.select(target_col).to_series().to_numpy()\n",
"\n",
"# 创建数据集\n",
"train_dataset = lgb.Dataset(X_train_np, label=y_train_np)\n",
"val_dataset = lgb.Dataset(X_val_np, label=y_val_np, reference=train_dataset)\n",
"\n",
"# 用于存储评估结果\n",
"evals_result = {}\n",
"\n",
"# 使用与原模型相同的参数重新训练\n",
"# 正确的三分法： ， ，
"# 添加早停：
"booster_with_eval = lgb.train(\n",
" MODEL_PARAMS,\n",
" train_dataset,\n",
" num_boost_round=MODEL_PARAMS.get(\"n_estimators\", 100),\n",
" valid_sets=[train_dataset, val_dataset],\n",
" valid_names=[\"train\", \"val\"],\n",
" callbacks=[\n",
" lgb.record_evaluation(evals_result),\n",
" lgb.early_stopping(stopping_rounds=100, verbose=True),\n",
" ],\n",
")\n",
"\n",
"print(\"训练完成，指标已收集\")\n",
"\n",
"# 获取指标名称\n",
"metric_name = list(evals_result[\"train\"].keys())[0]\n",
"print(f\"\\n评估指标: {metric_name}\")\n",
"\n",
"# 提取训练和验证指标\n",
"train_metric = evals_result[\"train\"][metric_name]\n",
"val_metric = evals_result[\"val\"][metric_name]\n",
"\n",
"# 显示早停信息\n",
"actual_rounds = len(train_metric)\n",
"expected_rounds = MODEL_PARAMS.get(\"n_estimators\", 100)\n",
"print(f\"\\n[早停信息]\")\n",
"print(f\" 配置的最大轮数: {expected_rounds}\")\n",
"print(f\" 实际训练轮数: {actual_rounds}\")\n",
"if actual_rounds < expected_rounds:\n",
" print(f\" 早停状态: 已触发（ ）
"else:\n",
" print(f\" 早停状态: 未触发（达到最大轮数）\")\n",
"\n",
"print(f\"\\n最终指标:\")\n",
"print(f\" 训练 {metric_name}: {train_metric[-1]:.6f}\")\n",
"print(f\" 验证 {metric_name}: {val_metric[-1]:.6f}\")"
],
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"================================================================================\n",
"训练指标曲线\n",
"================================================================================\n",
"\n",
"重新训练模型以收集训练指标...\n",
"Training until validation scores don't improve for 100 rounds\n",
"Early stopping, best iteration is:\n",
2026-03-09 23:37:20 +08:00
"[285]\ttrain's l1: 0.0410031\tval's l1: 0.0634246\n",
2026-03-08 01:09:47 +08:00
"训练完成，指标已收集\n",
"\n",
"评估指标: l1\n",
"\n",
"[早停信息]\n",
" 配置的最大轮数: 1000\n",
2026-03-09 23:37:20 +08:00
" 实际训练轮数: 385\n",
2026-03-08 01:09:47 +08:00
" 早停状态: 已触发（ ）
"\n",
"最终指标:\n",
2026-03-09 23:37:20 +08:00
" 训练 l1: 0.040855\n",
" 验证 l1: 0.063455\n"
2026-03-08 01:09:47 +08:00
]
}
],
2026-03-09 22:33:41 +08:00
"execution_count": 12
2026-03-08 01:09:47 +08:00
},
{
"cell_type": "code",
"metadata": {
"ExecuteTime": {
2026-03-09 23:37:20 +08:00
"end_time": "2026-03-09T15:27:15.271749Z",
"start_time": "2026-03-09T15:27:15.035816Z"
2026-03-08 01:09:47 +08:00
}
},
"source": [
"# 绘制训练指标曲线\n",
"import matplotlib.pyplot as plt\n",
"\n",
"fig, ax = plt.subplots(figsize=(12, 6))\n",
"\n",
"# 绘制训练集和验证集的指标曲线（ ： ， ）
"iterations = range(1, len(train_metric) + 1)\n",
"ax.plot(iterations, train_metric, label=f\"Train {metric_name}\", linewidth=2, color=\"blue\")\n",
"ax.plot(iterations, val_metric, label=f\"Validation {metric_name}\", linewidth=2, color=\"red\")\n",
"\n",
"ax.set_xlabel(\"Iteration\", fontsize=12)\n",
"ax.set_ylabel(metric_name.upper(), fontsize=12)\n",
"ax.set_title(f\"Training and Validation {metric_name.upper()} Curve\", fontsize=14, fontweight=\"bold\")\n",
"ax.legend(fontsize=10)\n",
"ax.grid(True, alpha=0.3)\n",
"\n",
"# 标记最佳验证指标点（用于早停决策）\n",
"best_iter = val_metric.index(min(val_metric))\n",
"best_metric = min(val_metric)\n",
"ax.axvline(x=best_iter + 1, color=\"green\", linestyle=\"--\", alpha=0.7, label=f\"Best Iteration ({best_iter + 1})\")\n",
"ax.scatter([best_iter + 1], [best_metric], color=\"green\", s=100, zorder=5)\n",
"ax.annotate(\n",
" f\"Best: {best_metric:.6f}\\nIter: {best_iter + 1}\",\n",
" xy=(best_iter + 1, best_metric),\n",
" xytext=(best_iter + 1 + len(iterations) * 0.1, best_metric),\n",
" fontsize=9,\n",
" arrowprops=dict(arrowstyle=\"->\", color=\"green\", alpha=0.7),\n",
")\n",
"\n",
"plt.tight_layout()\n",
"plt.show()\n",
"\n",
"print(f\"\\n[指标分析]\")\n",
"print(f\" 最佳验证 {metric_name}: {best_metric:.6f}\")\n",
"print(f\" 最佳迭代轮数: {best_iter + 1}\")\n",
"print(f\" 早停建议: 如果验证指标连续10轮不下降，
"print(f\"\\n[重要提醒] 验证集仅用于早停/调参，测试集完全独立于训练过程！\")"
],
"outputs": [
{
"data": {
"text/plain": [
"<Figure size 1200x600 with 1 Axes>"
],
2026-03-09 23:37:20 +08:00
"image/png": "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
2026-03-08 01:09:47 +08:00
},
"metadata": {},
"output_type": "display_data",
"jetTransient": {
"display_id": null
}
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"[指标分析]\n",
2026-03-09 23:37:20 +08:00
" 最佳验证 l1: 0.063425\n",
" 最佳迭代轮数: 285\n",
" 早停建议: 如果验证指标连续10轮不下降，
2026-03-08 01:09:47 +08:00
"\n",
"[重要提醒] 验证集仅用于早停/调参，测试集完全独立于训练过程！\n"
2026-03-06 20:57:27 +08:00
]
}
],
2026-03-09 22:33:41 +08:00
"execution_count": 13
2026-03-06 20:57:27 +08:00
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
2026-03-08 01:09:47 +08:00
"### 4.4 查看结果"
2026-03-06 20:57:27 +08:00
]
},
{
"cell_type": "code",
"metadata": {
"ExecuteTime": {
2026-03-09 23:37:20 +08:00
"end_time": "2026-03-09T15:27:15.311741Z",
"start_time": "2026-03-09T15:27:15.288652Z"
2026-03-06 20:57:27 +08:00
}
},
"source": [
"print(\"\\n\" + \"=\" * 80)\n",
"print(\"训练结果\")\n",
"print(\"=\" * 80)\n",
"\n",
"results = trainer.results\n",
"\n",
"print(f\"\\n结果数据形状: {results.shape}\")\n",
"print(f\"结果列: {results.columns}\")\n",
"print(f\"\\n结果前10行预览:\")\n",
"print(results.head(10))\n",
"print(f\"\\n结果后5行预览:\")\n",
"print(results.tail())\n",
"\n",
"print(f\"\\n每日预测样本数统计:\")\n",
"daily_counts = results.group_by(\"trade_date\").agg(pl.len()).sort(\"trade_date\")\n",
"print(f\" 最小: {daily_counts['len'].min()}\")\n",
"print(f\" 最大: {daily_counts['len'].max()}\")\n",
"print(f\" 平均: {daily_counts['len'].mean():.2f}\")\n",
"\n",
"# 展示某一天的前10个预测结果\n",
"sample_date = results[\"trade_date\"][0]\n",
"sample_data = results.filter(results[\"trade_date\"] == sample_date).head(10)\n",
"print(f\"\\n示例日期 {sample_date} 的前10条预测:\")\n",
"print(sample_data.select([\"ts_code\", \"trade_date\", target_col, \"prediction\"]))"
],
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"================================================================================\n",
"训练结果\n",
"================================================================================\n",
"\n",
2026-03-09 23:37:20 +08:00
"结果数据形状: (282000, 71)\n",
"结果列: ['ts_code', 'trade_date', 'high', 'low', 'amount', 'vol', 'close', 'turnover_rate', 'open', 'total_assets', 'total_mv', 'f_ann_date', 'total_cur_liab', 'total_cur_assets', 'total_hldr_eqy_exc_min_int', 'total_liab', 'revenue', 'n_income', 'n_cashflow_act', 'ebit', 'ma_5', 'ma_20', 'ma_ratio_5_20', 'bias_10', 'high_low_ratio', 'bbi_ratio', 'return_5', 'return_20', 'kaufman_ER_20', 'mom_acceleration_10_20', 'drawdown_from_high_60', 'up_days_ratio_20', 'volatility_5', 'volatility_20', 'volatility_ratio', 'std_return_20', 'sharpe_ratio_20', 'min_ret_20', 'volatility_squeeze_5_60', 'overnight_intraday_diff', 'upper_shadow_ratio', 'capital_retention_20', 'max_ret_20', 'volume_ratio_5_20', 'turnover_rate_mean_5', 'turnover_deviation', 'amihud_illiq_20', 'turnover_cv_20', 'pv_corr_20', 'close_vwap_deviation', 'roe', 'roa', 'profit_margin', 'debt_to_equity', 'current_ratio', 'net_profit_yoy', 'revenue_yoy', 'healthy_expansion_velocity', 'EP', 'BP', 'CP', 'market_cap_rank', 'turnover_rank', 'return_5_rank', 'EP_rank', 'pe_expansion_trend', 'value_price_divergence', 'active_market_cap', 'ebit_rank', 'future_return_5', 'prediction']\n",
2026-03-06 20:57:27 +08:00
"\n",
"结果前10行预览:\n",
2026-03-09 23:37:20 +08:00
"shape: (10, 71)\n",
"┌───────────┬────────────┬──────────┬──────────┬───┬───────────┬───────────┬───────────┬───────────┐\n",
"│ ts_code ┆ trade_date ┆ high ┆ low ┆ … ┆ active_ma ┆ ebit_rank ┆ future_re ┆ predictio │\n",
"│ --- ┆ --- ┆ --- ┆ --- ┆ ┆ rket_cap ┆ --- ┆ turn_5 ┆ n │\n",
"│ str ┆ str ┆ f64 ┆ f64 ┆ ┆ --- ┆ f64 ┆ --- ┆ --- │\n",
"│ ┆ ┆ ┆ ┆ ┆ f64 ┆ ┆ f64 ┆ f64 │\n",
"╞═══════════╪════════════╪══════════╪══════════╪═══╪═══════════╪═══════════╪═══════════╪═══════════╡\n",
"│ 000004.SZ ┆ 20250102 ┆ 2.325435 ┆ 2.168285 ┆ … ┆ 1.772242 ┆ null ┆ -0.066193 ┆ 0.023749 │\n",
"│ 000004.SZ ┆ 20250103 ┆ 2.328181 ┆ 2.003659 ┆ … ┆ 1.567863 ┆ null ┆ 0.00893 ┆ 0.041868 │\n",
"│ 000004.SZ ┆ 20250106 ┆ 1.918286 ┆ 1.811951 ┆ … ┆ 1.531265 ┆ null ┆ -0.0142 ┆ 0.054817 │\n",
"│ 000004.SZ ┆ 20250107 ┆ 1.971154 ┆ 1.971589 ┆ … ┆ 1.664744 ┆ null ┆ 0.013031 ┆ 0.025283 │\n",
"│ 000004.SZ ┆ 20250108 ┆ 2.071396 ┆ 2.040718 ┆ … ┆ 1.748053 ┆ null ┆ 0.00442 ┆ 0.013557 │\n",
"│ 000004.SZ ┆ 20250109 ┆ 2.085814 ┆ 2.134077 ┆ … ┆ 1.752624 ┆ null ┆ 0.024865 ┆ 0.005967 │\n",
"│ 000004.SZ ┆ 20250110 ┆ 2.03844 ┆ 1.928116 ┆ … ┆ 1.563176 ┆ null ┆ 0.073486 ┆ 0.012488 │\n",
"│ 000004.SZ ┆ 20250113 ┆ 1.800879 ┆ 1.745673 ┆ … ┆ 1.354823 ┆ null ┆ -0.04458 ┆ 0.010749 │\n",
"│ 000004.SZ ┆ 20250114 ┆ 1.993125 ┆ 1.936668 ┆ … ┆ 1.362504 ┆ null ┆ -0.152621 ┆ 0.004506 │\n",
"│ 000004.SZ ┆ 20250115 ┆ 2.188803 ┆ 2.154032 ┆ … ┆ 1.361079 ┆ null ┆ -0.152621 ┆ 0.005335 │\n",
"└───────────┴────────────┴──────────┴──────────┴───┴───────────┴───────────┴───────────┴───────────┘\n",
2026-03-06 20:57:27 +08:00
"\n",
"结果后5行预览:\n",
2026-03-09 23:37:20 +08:00
"shape: (5, 71)\n",
"┌───────────┬────────────┬──────────┬──────────┬───┬───────────┬───────────┬───────────┬───────────┐\n",
"│ ts_code ┆ trade_date ┆ high ┆ low ┆ … ┆ active_ma ┆ ebit_rank ┆ future_re ┆ predictio │\n",
"│ --- ┆ --- ┆ --- ┆ --- ┆ ┆ rket_cap ┆ --- ┆ turn_5 ┆ n │\n",
"│ str ┆ str ┆ f64 ┆ f64 ┆ ┆ --- ┆ f64 ┆ --- ┆ --- │\n",
"│ ┆ ┆ ┆ ┆ ┆ f64 ┆ ┆ f64 ┆ f64 │\n",
"╞═══════════╪════════════╪══════════╪══════════╪═══╪═══════════╪═══════════╪═══════════╪═══════════╡\n",
"│ 605588.SH ┆ 20260302 ┆ 2.566428 ┆ 2.53816 ┆ … ┆ 0.112914 ┆ null ┆ null ┆ -0.002814 │\n",
"│ 605588.SH ┆ 20260303 ┆ 2.475112 ┆ 2.280174 ┆ … ┆ 0.061385 ┆ null ┆ null ┆ 0.021067 │\n",
"│ 605588.SH ┆ 20260304 ┆ 2.173698 ┆ 2.114835 ┆ … ┆ 0.028725 ┆ null ┆ null ┆ 0.007579 │\n",
"│ 605588.SH ┆ 20260305 ┆ 2.18537 ┆ 2.219597 ┆ … ┆ 0.044869 ┆ null ┆ null ┆ 0.006004 │\n",
"│ 605588.SH ┆ 20260306 ┆ 2.171638 ┆ 2.201068 ┆ … ┆ 0.021626 ┆ null ┆ null ┆ 0.004578 │\n",
"└───────────┴────────────┴──────────┴──────────┴───┴───────────┴───────────┴───────────┴───────────┘\n",
2026-03-06 20:57:27 +08:00
"\n",
"每日预测样本数统计:\n",
2026-03-09 22:33:41 +08:00
" 最小: 1000\n",
" 最大: 1000\n",
" 平均: 1000.00\n",
2026-03-06 20:57:27 +08:00
"\n",
"示例日期 20250102 的前10条预测:\n",
"shape: (10, 4)\n",
2026-03-08 14:12:03 +08:00
"┌───────────┬────────────┬─────────────────┬────────────┐\n",
"│ ts_code ┆ trade_date ┆ future_return_5 ┆ prediction │\n",
"│ --- ┆ --- ┆ --- ┆ --- │\n",
"│ str ┆ str ┆ f64 ┆ f64 │\n",
"╞═══════════╪════════════╪═════════════════╪════════════╡\n",
2026-03-09 23:37:20 +08:00
"│ 000004.SZ ┆ 20250102 ┆ -0.066193 ┆ 0.023749 │\n",
"│ 000007.SZ ┆ 20250102 ┆ 0.019858 ┆ 0.000582 │\n",
"│ 000010.SZ ┆ 20250102 ┆ 0.076274 ┆ 0.004461 │\n",
"│ 000014.SZ ┆ 20250102 ┆ -0.064651 ┆ 0.006334 │\n",
"│ 000040.SZ ┆ 20250102 ┆ -0.093583 ┆ -0.102073 │\n",
"│ 000042.SZ ┆ 20250102 ┆ -0.035958 ┆ 0.017509 │\n",
"│ 000056.SZ ┆ 20250102 ┆ -0.033205 ┆ 0.020625 │\n",
"│ 000068.SZ ┆ 20250102 ┆ -0.021277 ┆ 0.00916 │\n",
"│ 000153.SZ ┆ 20250102 ┆ -0.018193 ┆ 0.003244 │\n",
"│ 000159.SZ ┆ 20250102 ┆ -0.067833 ┆ 0.020453 │\n",
2026-03-08 14:12:03 +08:00
"└───────────┴────────────┴─────────────────┴────────────┘\n"
2026-03-06 20:57:27 +08:00
]
}
],
2026-03-09 22:33:41 +08:00
"execution_count": 14
2026-03-06 20:57:27 +08:00
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### 4.4 保存结果"
]
},
{
"metadata": {
"ExecuteTime": {
2026-03-09 23:37:20 +08:00
"end_time": "2026-03-09T15:27:15.611776Z",
"start_time": "2026-03-09T15:27:15.317283Z"
2026-03-06 20:57:27 +08:00
}
},
"cell_type": "code",
"source": [
"print(\"\\n\" + \"=\" * 80)\n",
"print(\"保存预测结果\")\n",
"print(\"=\" * 80)\n",
"\n",
"# 确保输出目录存在\n",
"os.makedirs(OUTPUT_DIR, exist_ok=True)\n",
"\n",
"# 生成时间戳\n",
"start_dt = datetime.strptime(TEST_START, \"%Y%m%d\")\n",
"end_dt = datetime.strptime(TEST_END, \"%Y%m%d\")\n",
"date_str = f\"{start_dt.strftime('%Y%m%d')}_{end_dt.strftime('%Y%m%d')}\"\n",
"\n",
2026-03-08 11:46:30 +08:00
"# 保存每日 Top N\n",
"print(f\"\\n[1/1] 保存每日 Top {TOP_N} 股票...\")\n",
"topn_output_path = os.path.join(OUTPUT_DIR, f\"regression_output.csv\")\n",
2026-03-06 20:57:27 +08:00
"\n",
2026-03-08 11:46:30 +08:00
"# 按日期分组，取每日 top N\n",
"topn_by_date = []\n",
2026-03-06 20:57:27 +08:00
"unique_dates = results[\"trade_date\"].unique().sort()\n",
"for date in unique_dates:\n",
" day_data = results.filter(results[\"trade_date\"] == date)\n",
2026-03-08 11:46:30 +08:00
" # 按 prediction 降序排序，取前 N\n",
" topn = day_data.sort(\"prediction\", descending=True).head(TOP_N)\n",
" topn_by_date.append(topn)\n",
2026-03-06 20:57:27 +08:00
"\n",
2026-03-08 11:46:30 +08:00
"# 合并所有日期的 top N\n",
"topn_results = pl.concat(topn_by_date)\n",
2026-03-06 20:57:27 +08:00
"\n",
"# 格式化日期并调整列顺序：日期、分数、股票\n",
2026-03-08 11:46:30 +08:00
"topn_to_save = topn_results.select(\n",
2026-03-06 20:57:27 +08:00
" [\n",
" pl.col(\"trade_date\").str.slice(0, 4)\n",
" + \"-\"\n",
" + pl.col(\"trade_date\").str.slice(4, 2)\n",
" + \"-\"\n",
" + pl.col(\"trade_date\").str.slice(6, 2).alias(\"date\"),\n",
" pl.col(\"prediction\").alias(\"score\"),\n",
" pl.col(\"ts_code\"),\n",
" ]\n",
")\n",
2026-03-08 11:46:30 +08:00
"topn_to_save.write_csv(topn_output_path, include_header=True)\n",
"print(f\" 保存路径: {topn_output_path}\")\n",
"print(f\" 保存行数: {len(topn_to_save)}（ × ）
2026-03-06 20:57:27 +08:00
"print(f\"\\n 预览（ ）
2026-03-08 11:46:30 +08:00
"print(topn_to_save.head(15))"
2026-03-06 20:57:27 +08:00
],
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"================================================================================\n",
"保存预测结果\n",
"================================================================================\n",
"\n",
2026-03-08 20:58:35 +08:00
"[1/1] 保存每日 Top 5 股票...\n",
2026-03-08 11:46:30 +08:00
" 保存路径: output\\regression_output.csv\n",
2026-03-09 22:33:41 +08:00
" 保存行数: 1410（ × ）
2026-03-06 20:57:27 +08:00
"\n",
" 预览（ ）
"shape: (15, 3)\n",
"┌────────────┬──────────┬───────────┐\n",
"│ trade_date ┆ score ┆ ts_code │\n",
"│ --- ┆ --- ┆ --- │\n",
"│ str ┆ f64 ┆ str │\n",
"╞════════════╪══════════╪═══════════╡\n",
2026-03-09 23:37:20 +08:00
"│ 2025-01-02 ┆ 0.056168 ┆ 600421.SH │\n",
"│ 2025-01-02 ┆ 0.048049 ┆ 000668.SZ │\n",
"│ 2025-01-02 ┆ 0.042422 ┆ 000586.SZ │\n",
"│ 2025-01-02 ┆ 0.036414 ┆ 002076.SZ │\n",
"│ 2025-01-02 ┆ 0.035977 ┆ 301176.SZ │\n",
2026-03-06 20:57:27 +08:00
"│ … ┆ … ┆ … │\n",
2026-03-09 23:37:20 +08:00
"│ 2025-01-06 ┆ 0.069458 ┆ 600421.SH │\n",
"│ 2025-01-06 ┆ 0.067984 ┆ 603316.SH │\n",
"│ 2025-01-06 ┆ 0.062674 ┆ 301024.SZ │\n",
"│ 2025-01-06 ┆ 0.062283 ┆ 002691.SZ │\n",
"│ 2025-01-06 ┆ 0.060667 ┆ 000668.SZ │\n",
2026-03-06 20:57:27 +08:00
"└────────────┴──────────┴───────────┘\n"
]
}
],
2026-03-09 22:33:41 +08:00
"execution_count": 15
2026-03-06 20:57:27 +08:00
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### 4.5 特征重要性"
]
},
{
"cell_type": "code",
"metadata": {
"ExecuteTime": {
2026-03-09 23:37:20 +08:00
"end_time": "2026-03-09T15:27:15.622544Z",
"start_time": "2026-03-09T15:27:15.617560Z"
2026-03-06 20:57:27 +08:00
}
},
"source": [
"importance = model.feature_importance()\n",
"if importance is not None:\n",
" print(\"\\n特征重要性:\")\n",
" print(importance.sort_values(ascending=False))\n",
"\n",
"print(\"\\n\" + \"=\" * 80)\n",
"print(\"训练完成！\")\n",
"print(\"=\" * 80)"
],
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"特征重要性:\n",
2026-03-09 23:37:20 +08:00
"bias_10 966.502486\n",
"return_5_rank 960.507971\n",
"pe_expansion_trend 947.437883\n",
"return_5 860.686048\n",
"high_low_ratio 727.172214\n",
"revenue_yoy 562.656819\n",
"turnover_rank 558.328422\n",
"amihud_illiq_20 388.591107\n",
"overnight_intraday_diff 380.723414\n",
"bbi_ratio 373.573519\n",
"min_ret_20 366.693362\n",
"drawdown_from_high_60 356.182625\n",
"roa 334.564408\n",
"active_market_cap 275.635468\n",
"ma_ratio_5_20 261.307464\n",
"turnover_deviation 194.988921\n",
"return_20 192.620460\n",
"net_profit_yoy 175.035165\n",
"mom_acceleration_10_20 171.881004\n",
"turnover_rate_mean_5 168.826981\n",
"EP_rank 163.394623\n",
"volume_ratio_5_20 158.494889\n",
"turnover_cv_20 153.915966\n",
"healthy_expansion_velocity 146.708935\n",
"EP 142.683564\n",
"max_ret_20 135.638350\n",
"std_return_20 133.794949\n",
"sharpe_ratio_20 108.012163\n",
"close_vwap_deviation 103.748435\n",
"ma_20 80.322421\n",
"volatility_squeeze_5_60 77.509971\n",
"BP 64.954998\n",
"ma_5 60.884566\n",
"roe 59.556815\n",
"volatility_ratio 53.867742\n",
"capital_retention_20 45.922123\n",
"volatility_5 41.385944\n",
"volatility_20 34.698196\n",
"profit_margin 29.828428\n",
"debt_to_equity 28.911864\n",
"pv_corr_20 27.893618\n",
"current_ratio 22.564712\n",
"upper_shadow_ratio 16.570660\n",
"kaufman_ER_20 16.416594\n",
"market_cap_rank 15.822645\n",
"CP 13.325475\n",
"up_days_ratio_20 10.937868\n",
"value_price_divergence 0.000000\n",
"ebit_rank 0.000000\n",
2026-03-06 20:57:27 +08:00
"dtype: float64\n",
"\n",
"================================================================================\n",
"训练完成！\n",
"================================================================================\n"
]
}
],
2026-03-09 22:33:41 +08:00
"execution_count": 16
2026-03-06 20:57:27 +08:00
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## 5. 可视化分析\n",
"\n",
"使用训练好的模型直接绘图。\n",
"- **特征重要性图**：辅助特征选择\n",
"- **决策树图**：理解决策逻辑"
]
},
{
"cell_type": "code",
"metadata": {
"ExecuteTime": {
2026-03-09 23:37:20 +08:00
"end_time": "2026-03-09T15:27:15.634579Z",
"start_time": "2026-03-09T15:27:15.631711Z"
2026-03-06 20:57:27 +08:00
}
},
"source": [
"# 导入可视化库\n",
"import matplotlib.pyplot as plt\n",
"import lightgbm as lgb\n",
"import pandas as pd\n",
"\n",
"# 从封装的model中取出底层Booster\n",
"booster = model.model\n",
"print(f\"模型类型: {type(booster)}\")\n",
"print(f\"特征数量: {len(feature_cols)}\")"
],
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"模型类型: <class 'lightgbm.basic.Booster'>\n",
2026-03-09 23:37:20 +08:00
"特征数量: 49\n"
2026-03-06 20:57:27 +08:00
]
}
],
2026-03-09 22:33:41 +08:00
"execution_count": 17
2026-03-06 20:57:27 +08:00
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### 5.1 绘制特征重要性（辅助特征选择）\n",
"\n",
"**解读**：
"- 重要性高的特征对模型贡献大\n",
"- 重要性为0的特征可以考虑删除\n",
"- 可以帮助理解哪些因子最有效"
]
},
{
"metadata": {
"ExecuteTime": {
2026-03-09 23:37:20 +08:00
"end_time": "2026-03-09T15:27:15.796755Z",
"start_time": "2026-03-09T15:27:15.641538Z"
2026-03-06 20:57:27 +08:00
}
},
"cell_type": "code",
"source": [
"print(\"绘制特征重要性...\")\n",
"\n",
"fig, ax = plt.subplots(figsize=(10, 8))\n",
"lgb.plot_importance(\n",
2026-03-09 22:33:41 +08:00
" booster,\n",
2026-03-06 20:57:27 +08:00
" max_num_features=20,\n",
" importance_type='gain',\n",
" title='Feature Importance (Gain)',\n",
" ax=ax\n",
")\n",
"ax.set_xlabel('Importance (Gain)')\n",
"plt.tight_layout()\n",
"plt.show()\n",
"\n",
"# 打印重要性排名\n",
"importance_gain = pd.Series(\n",
" booster.feature_importance(importance_type='gain'),\n",
" index=feature_cols\n",
").sort_values(ascending=False)\n",
"\n",
"print(\"\\n[特征重要性排名 - Gain]\")\n",
"print(importance_gain)\n",
"\n",
"# 识别低重要性特征\n",
"zero_importance = importance_gain[importance_gain == 0].index.tolist()\n",
"if zero_importance:\n",
" print(f\"\\n[低重要性特征] 以下{len(zero_importance)}个特征重要性为0，
" for feat in zero_importance:\n",
" print(f\" - {feat}\")\n",
"else:\n",
" print(\"\\n所有特征都有一定重要性\")"
],
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"绘制特征重要性...\n"
]
},
{
"data": {
"text/plain": [
"<Figure size 1000x800 with 1 Axes>"
],
2026-03-09 23:37:20 +08:00
"image/png": "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
2026-03-06 20:57:27 +08:00
},
"metadata": {},
"output_type": "display_data",
"jetTransient": {
"display_id": null
}
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"[特征重要性排名 - Gain]\n",
2026-03-09 23:37:20 +08:00
"bias_10 966.502486\n",
"return_5_rank 960.507971\n",
"pe_expansion_trend 947.437883\n",
"return_5 860.686048\n",
"high_low_ratio 727.172214\n",
"revenue_yoy 562.656819\n",
"turnover_rank 558.328422\n",
"amihud_illiq_20 388.591107\n",
"overnight_intraday_diff 380.723414\n",
"bbi_ratio 373.573519\n",
"min_ret_20 366.693362\n",
"drawdown_from_high_60 356.182625\n",
"roa 334.564408\n",
"active_market_cap 275.635468\n",
"ma_ratio_5_20 261.307464\n",
"turnover_deviation 194.988921\n",
"return_20 192.620460\n",
"net_profit_yoy 175.035165\n",
"mom_acceleration_10_20 171.881004\n",
"turnover_rate_mean_5 168.826981\n",
"EP_rank 163.394623\n",
"volume_ratio_5_20 158.494889\n",
"turnover_cv_20 153.915966\n",
"healthy_expansion_velocity 146.708935\n",
"EP 142.683564\n",
"max_ret_20 135.638350\n",
"std_return_20 133.794949\n",
"sharpe_ratio_20 108.012163\n",
"close_vwap_deviation 103.748435\n",
"ma_20 80.322421\n",
"volatility_squeeze_5_60 77.509971\n",
"BP 64.954998\n",
"ma_5 60.884566\n",
"roe 59.556815\n",
"volatility_ratio 53.867742\n",
"capital_retention_20 45.922123\n",
"volatility_5 41.385944\n",
"volatility_20 34.698196\n",
"profit_margin 29.828428\n",
"debt_to_equity 28.911864\n",
"pv_corr_20 27.893618\n",
"current_ratio 22.564712\n",
"upper_shadow_ratio 16.570660\n",
"kaufman_ER_20 16.416594\n",
"market_cap_rank 15.822645\n",
"CP 13.325475\n",
"up_days_ratio_20 10.937868\n",
"value_price_divergence 0.000000\n",
"ebit_rank 0.000000\n",
2026-03-06 20:57:27 +08:00
"dtype: float64\n",
"\n",
2026-03-09 22:33:41 +08:00
"[低重要性特征] 以下2个特征重要性为0，
2026-03-09 23:37:20 +08:00
" - value_price_divergence\n",
" - ebit_rank\n"
2026-03-06 20:57:27 +08:00
]
}
],
2026-03-09 22:33:41 +08:00
"execution_count": 18
2026-03-06 20:57:27 +08:00
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.10.0"
}
},
"nbformat": 4,
"nbformat_minor": 4
}
