{ "cells": [ { "cell_type": "code", "execution_count": 1, "id": "782ec73f", "metadata": {}, "outputs": [], "source": [ "import pandas as pd\n", "from datetime import datetime\n", "import itertools\n", "from typing import Dict, Any, List, Tuple, Optional\n", "import multiprocessing # 导入 multiprocessing 模块\n", "import math # 保留 math 导入，因为您的策略内部可能需要用到数学函数\n", "\n", "# 导入所有必要的模块\n", "# 请确保这些导入路径与您的项目结构相符\n", "from src.analysis.grid_search_analyzer import GridSearchAnalyzer\n", "from src.analysis.result_analyzer import ResultAnalyzer\n", "from src.common_utils import generate_parameter_range\n", "from src.data_manager import DataManager\n", "from src.backtest_engine import BacktestEngine\n", "# 导入策略类\n", "from src.strategies.SimpleLimitBuyStrategy import SimpleLimitBuyStrategyShort, SimpleLimitBuyStrategyLong, SimpleLimitBuyStrategy\n", "\n", "\n", "import builtins\n", "\n", "%load_ext autoreload\n", "%autoreload 2\n", "\n", "origin_print = print\n" ] }, { "cell_type": "code", "execution_count": 2, "id": "76f9a2e9", "metadata": {}, "outputs": [], "source": [ "\n", "# --- 单个回测任务函数 ---\n", "# 这个函数将在每个独立的进程中运行，因此它必须是自包含的\n", "def run_single_backtest(\n", " combination: Tuple[float, float], # 传入当前参数组合\n", " common_config: Dict[str, Any] # 传入公共配置 (如数据路径, 初始资金等)\n", ") -> Optional[Dict[str, Any]]:\n", " \"\"\"\n", " 运行单个参数组合的回测任务。\n", " 此函数将在一个独立的进程中执行。\n", " \"\"\"\n", " p1_value, p2_value = combination\n", "\n", " # 从 common_config 中获取必要的配置\n", " symbol = common_config['symbol']\n", " data_path = common_config['data_path']\n", " initial_capital = common_config['initial_capital']\n", " slippage_rate = common_config['slippage_rate']\n", " commission_rate = common_config['commission_rate']\n", " start_time = common_config['start_time']\n", " end_time = common_config['end_time']\n", " roll_over_mode = common_config['roll_over_mode']\n", " # bar_duration_seconds = common_config['bar_duration_seconds'] # 如果DataManager需要，可以再传\n", " param1_name = common_config['param1_name']\n", " param2_name = common_config['param2_name']\n", " optimization_metric = common_config['optimization_metric']\n", "\n", " # 每个进程内部独立初始化 DataManager 和 BacktestEngine\n", " # 确保每个进程有自己的数据副本和模拟状态\n", " data_manager = DataManager(\n", " file_path=data_path,\n", " symbol=symbol,\n", " # bar_duration_seconds=bar_duration_seconds, # 如果DataManager需要，根据数据文件路径推断或者额外参数传入\n", " # start_date=start_time.date(), # DataManager 现在通过 file_path 和 symbol 处理数据\n", " # end_date=end_time.date(),\n", " )\n", " # data_manager.load_data() # DataManager 内部加载数据\n", "\n", " # 策略参数\n", " strategy_parameters = {\n", " 'trade_volume': 1,\n", " param1_name: p1_value,\n", " param2_name: p2_value,\n", " 'max_position': 10,\n", " 'enable_log': False, # 在网格搜索时通常关闭策略内部的详细日志\n", " }\n", "\n", " # 打印当前进程正在处理的组合信息\n", " # 注意：多进程打印会交错显示\n", " # print(f\"--- 正在运行组合: {strategy_parameters} (PID: {multiprocessing.current_process().pid}) ---\")\n", "\n", " try:\n", " # 初始化回测引擎\n", " engine = BacktestEngine(\n", " data_manager=data_manager,\n", " strategy_class=common_config['strategy'],\n", " strategy_params=strategy_parameters,\n", " initial_capital=initial_capital,\n", " slippage_rate=slippage_rate,\n", " commission_rate=commission_rate,\n", " roll_over_mode=True, # 保持换月模式\n", " start_time=common_config['start_time'],\n", " end_time=common_config['end_time']\n", " )\n", " # 运行回测，传入时间范围\n", " engine.run_backtest()\n", "\n", " # 获取回测结果并分析\n", " results = engine.get_backtest_results()\n", " portfolio_snapshots = results[\"portfolio_snapshots\"]\n", " trade_history = results[\"trade_history\"]\n", " bars = results[\"all_bars\"]\n", " initial_capital_result = results[\"initial_capital\"]\n", "\n", " if portfolio_snapshots:\n", " analyzer = ResultAnalyzer(portfolio_snapshots, trade_history, bars, initial_capital_result)\n", "\n", " # analyzer.generate_report()\n", " # analyzer.plot_performance()\n", " metrics = analyzer.calculate_all_metrics()\n", "\n", " # 将当前组合的参数和性能指标存储起来\n", " result_entry = {**strategy_parameters, **metrics}\n", " # print(f\" 组合 {combination} 完成。{optimization_metric}: {metrics.get(optimization_metric, 0.0):.4f} (PID: {multiprocessing.current_process().pid})\")\n", " return result_entry\n", " else:\n", " print(f\" 组合 {strategy_parameters} 没有生成投资组合快照，无法进行结果分析。(PID: {multiprocessing.current_process().pid})\")\n", " # 返回一个包含参数和默认0值的结果，以便追踪失败组合\n", " return {**strategy_parameters, \"total_return\": 0.0, \"annualized_return\": 0.0, \"sharpe_ratio\": 0.0, \"max_drawdown\": 0.0, \"error\": \"No portfolio snapshots\"}\n", " except Exception as e:\n", " import traceback\n", " error_trace = traceback.format_exc()\n", " print(f\" 组合 {strategy_parameters} 运行失败: {e}\\n{error_trace} (PID: {multiprocessing.current_process().pid})\")\n", " # 返回错误信息，以便后续处理\n", " return {**strategy_parameters, \"error\": str(e), \"traceback\": error_trace}\n", "\n" ] }, { "cell_type": "code", "execution_count": 3, "id": "c0984689", "metadata": {}, "outputs": [], "source": [ "\n", "def slient_print(*args):\n", " pass\n" ] }, { "cell_type": "code", "execution_count": 4, "id": "239e9ca0", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "总计 1681 种参数组合需要回测。\n", "--- 启动多进程网格搜索，使用 10 个进程 ---\n" ] }, { "ename": "KeyboardInterrupt", "evalue": "", "output_type": "error", "traceback": [ "\u001b[31m---------------------------------------------------------------------------\u001b[39m", "\u001b[31mKeyboardInterrupt\u001b[39m Traceback (most recent call last)", "\u001b[36mCell\u001b[39m\u001b[36m \u001b[39m\u001b[32mIn[4]\u001b[39m\u001b[32m, line 69\u001b[39m\n\u001b[32m 63\u001b[39m args_for_starmap = [\n\u001b[32m 64\u001b[39m (combo, common_config_for_processes) \u001b[38;5;28;01mfor\u001b[39;00m combo \u001b[38;5;129;01min\u001b[39;00m param_combinations\n\u001b[32m 65\u001b[39m ]\n\u001b[32m 67\u001b[39m \u001b[38;5;66;03m# 使用 starmap() 来并行执行 run_single_backtest 函数\u001b[39;00m\n\u001b[32m 68\u001b[39m \u001b[38;5;66;03m# starmap 是阻塞的，会等待所有任务完成并返回结果列表\u001b[39;00m\n\u001b[32m---> \u001b[39m\u001b[32m69\u001b[39m \u001b[38;5;28;01mfor\u001b[39;00m i, result_entry \u001b[38;5;129;01min\u001b[39;00m \u001b[38;5;28menumerate\u001b[39m(\u001b[43mpool\u001b[49m\u001b[43m.\u001b[49m\u001b[43mstarmap\u001b[49m\u001b[43m(\u001b[49m\u001b[43mrun_single_backtest\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43margs_for_starmap\u001b[49m\u001b[43m)\u001b[49m):\n\u001b[32m 70\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m result_entry: \u001b[38;5;66;03m# 确保结果不为空\u001b[39;00m\n\u001b[32m 71\u001b[39m all_results.append(result_entry)\n", "\u001b[36mFile \u001b[39m\u001b[32m~/miniconda3/envs/quant/lib/python3.12/multiprocessing/pool.py:375\u001b[39m, in \u001b[36mPool.starmap\u001b[39m\u001b[34m(self, func, iterable, chunksize)\u001b[39m\n\u001b[32m 369\u001b[39m \u001b[38;5;28;01mdef\u001b[39;00m\u001b[38;5;250m \u001b[39m\u001b[34mstarmap\u001b[39m(\u001b[38;5;28mself\u001b[39m, func, iterable, chunksize=\u001b[38;5;28;01mNone\u001b[39;00m):\n\u001b[32m 370\u001b[39m \u001b[38;5;250m \u001b[39m\u001b[33;03m'''\u001b[39;00m\n\u001b[32m 371\u001b[39m \u001b[33;03m Like `map()` method but the elements of the `iterable` are expected to\u001b[39;00m\n\u001b[32m 372\u001b[39m \u001b[33;03m be iterables as well and will be unpacked as arguments. Hence\u001b[39;00m\n\u001b[32m 373\u001b[39m \u001b[33;03m `func` and (a, b) becomes func(a, b).\u001b[39;00m\n\u001b[32m 374\u001b[39m \u001b[33;03m '''\u001b[39;00m\n\u001b[32m--> \u001b[39m\u001b[32m375\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28;43mself\u001b[39;49m\u001b[43m.\u001b[49m\u001b[43m_map_async\u001b[49m\u001b[43m(\u001b[49m\u001b[43mfunc\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43miterable\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mstarmapstar\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mchunksize\u001b[49m\u001b[43m)\u001b[49m\u001b[43m.\u001b[49m\u001b[43mget\u001b[49m\u001b[43m(\u001b[49m\u001b[43m)\u001b[49m\n", "\u001b[36mFile \u001b[39m\u001b[32m~/miniconda3/envs/quant/lib/python3.12/multiprocessing/pool.py:768\u001b[39m, in \u001b[36mApplyResult.get\u001b[39m\u001b[34m(self, timeout)\u001b[39m\n\u001b[32m 767\u001b[39m \u001b[38;5;28;01mdef\u001b[39;00m\u001b[38;5;250m \u001b[39m\u001b[34mget\u001b[39m(\u001b[38;5;28mself\u001b[39m, timeout=\u001b[38;5;28;01mNone\u001b[39;00m):\n\u001b[32m--> \u001b[39m\u001b[32m768\u001b[39m \u001b[38;5;28;43mself\u001b[39;49m\u001b[43m.\u001b[49m\u001b[43mwait\u001b[49m\u001b[43m(\u001b[49m\u001b[43mtimeout\u001b[49m\u001b[43m)\u001b[49m\n\u001b[32m 769\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m \u001b[38;5;28mself\u001b[39m.ready():\n\u001b[32m 770\u001b[39m \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mTimeoutError\u001b[39;00m\n", "\u001b[36mFile \u001b[39m\u001b[32m~/miniconda3/envs/quant/lib/python3.12/multiprocessing/pool.py:765\u001b[39m, in \u001b[36mApplyResult.wait\u001b[39m\u001b[34m(self, timeout)\u001b[39m\n\u001b[32m 764\u001b[39m \u001b[38;5;28;01mdef\u001b[39;00m\u001b[38;5;250m \u001b[39m\u001b[34mwait\u001b[39m(\u001b[38;5;28mself\u001b[39m, timeout=\u001b[38;5;28;01mNone\u001b[39;00m):\n\u001b[32m--> \u001b[39m\u001b[32m765\u001b[39m \u001b[38;5;28;43mself\u001b[39;49m\u001b[43m.\u001b[49m\u001b[43m_event\u001b[49m\u001b[43m.\u001b[49m\u001b[43mwait\u001b[49m\u001b[43m(\u001b[49m\u001b[43mtimeout\u001b[49m\u001b[43m)\u001b[49m\n", "\u001b[36mFile \u001b[39m\u001b[32m~/miniconda3/envs/quant/lib/python3.12/threading.py:655\u001b[39m, in \u001b[36mEvent.wait\u001b[39m\u001b[34m(self, timeout)\u001b[39m\n\u001b[32m 653\u001b[39m signaled = \u001b[38;5;28mself\u001b[39m._flag\n\u001b[32m 654\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m signaled:\n\u001b[32m--> \u001b[39m\u001b[32m655\u001b[39m signaled = \u001b[38;5;28;43mself\u001b[39;49m\u001b[43m.\u001b[49m\u001b[43m_cond\u001b[49m\u001b[43m.\u001b[49m\u001b[43mwait\u001b[49m\u001b[43m(\u001b[49m\u001b[43mtimeout\u001b[49m\u001b[43m)\u001b[49m\n\u001b[32m 656\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m signaled\n", "\u001b[36mFile \u001b[39m\u001b[32m~/miniconda3/envs/quant/lib/python3.12/threading.py:355\u001b[39m, in \u001b[36mCondition.wait\u001b[39m\u001b[34m(self, timeout)\u001b[39m\n\u001b[32m 353\u001b[39m \u001b[38;5;28;01mtry\u001b[39;00m: \u001b[38;5;66;03m# restore state no matter what (e.g., KeyboardInterrupt)\u001b[39;00m\n\u001b[32m 354\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m timeout \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m:\n\u001b[32m--> \u001b[39m\u001b[32m355\u001b[39m \u001b[43mwaiter\u001b[49m\u001b[43m.\u001b[49m\u001b[43macquire\u001b[49m\u001b[43m(\u001b[49m\u001b[43m)\u001b[49m\n\u001b[32m 356\u001b[39m gotit = \u001b[38;5;28;01mTrue\u001b[39;00m\n\u001b[32m 357\u001b[39m \u001b[38;5;28;01melse\u001b[39;00m:\n", "\u001b[31mKeyboardInterrupt\u001b[39m: " ] } ], "source": [ "\n", "# --- 主执行块 ---\n", "# 这是多进程代码的入口点，必须在 'if __name__ == \"__main__\":' 保护块中\n", "if __name__ == \"__main__\":\n", " # 确保 autoreload 启用 (在Jupyter Notebook中使用，纯Python脚本运行时可移除)\n", " # %load_ext autoreload\n", " # %autoreload 2\n", "\n", " # --- 全局配置 ---\n", " data_file_path = \"/mnt/d/PyProject/NewQuant/data/data/KQ_m@CZCE_MA/KQ_m@CZCE_MA_min60.csv\"\n", " initial_capital = 100000.0\n", " slippage_rate = 0.0000\n", " commission_rate = 0.0002\n", " global_config = {\n", " 'symbol': 'KQ_m@CZCE_MA',\n", " }\n", " # 确保每个合约的tick_size在这里定义或获取\n", " RB_TICK_SIZE = 1.0 # 螺纹钢的最小变动单位\n", "\n", " # --- 定义参数网格 ---\n", " param1_name = \"open_range_factor_1_ago\"\n", " param1_values = generate_parameter_range(start=-2, end=2, step=0.1)\n", " param2_name = \"open_range_factor_7_ago\"\n", " param2_values = generate_parameter_range(start=-2, end=2, step=0.1)\n", " optimization_metric = 'sharpe_ratio'\n", " \n", " # 生成所有参数组合\n", " param_combinations = list(itertools.product(param1_values, param2_values))\n", " total_combinations = len(param_combinations)\n", " print(f\"总计 {total_combinations} 种参数组合需要回测。\")\n", "\n", " all_results: List[Dict[str, Any]] = []\n", " grid_results: List[Dict[str, Any]] = []\n", "\n", " # 准备传递给每个子进程的公共配置字典\n", " common_config_for_processes = {\n", " 'symbol': global_config['symbol'],\n", " 'data_path': data_file_path,\n", " 'initial_capital': initial_capital,\n", " 'slippage_rate': slippage_rate,\n", " 'commission_rate': commission_rate,\n", " 'start_time': datetime(2022, 1, 1), # 回测起始时间\n", " 'end_time': datetime(2025, 1, 1), # 回测结束时间\n", " 'roll_over_mode': True, # 保持换月模式\n", " 'param1_name': param1_name,\n", " 'param2_name': param2_name,\n", " 'optimization_metric': optimization_metric,\n", " 'strategy': SimpleLimitBuyStrategyLong\n", " }\n", "\n", " # 确定要使用的进程数量 (通常是CPU核心数)\n", " num_processes = int(multiprocessing.cpu_count() / 2)\n", " if num_processes < 1:\n", " num_processes = 1\n", " \n", " print(f\"--- 启动多进程网格搜索，使用 {num_processes} 个进程 ---\")\n", "\n", " builtins.print = slient_print\n", "\n", " # 创建一个进程池\n", " with multiprocessing.Pool(processes=num_processes) as pool:\n", " # 准备 run_single_backtest 函数的参数列表\n", " # starmap 需要一个可迭代对象，其中每个元素是传递给目标函数的参数元组\n", " args_for_starmap = [\n", " (combo, common_config_for_processes) for combo in param_combinations\n", " ]\n", " \n", " # 使用 starmap() 来并行执行 run_single_backtest 函数\n", " # starmap 是阻塞的，会等待所有任务完成并返回结果列表\n", " for i, result_entry in enumerate(pool.starmap(run_single_backtest, args_for_starmap)):\n", " if result_entry: # 确保结果不为空\n", " all_results.append(result_entry)\n", " # 仅将成功的（无错误的）结果添加到用于网格分析的列表中\n", " if 'error' not in result_entry:\n", " grid_results.append(\n", " {\n", " param1_name: result_entry.get(param1_name),\n", " param2_name: result_entry.get(param2_name),\n", " optimization_metric: result_entry.get(optimization_metric, 0.0),\n", " }\n", " )\n", " else:\n", " # 对于失败的组合，将其优化指标设置为一个特殊值，便于识别\n", " grid_results.append(\n", " {\n", " param1_name: result_entry.get(param1_name),\n", " param2_name: result_entry.get(param2_name),\n", " optimization_metric: float('-inf'), # 用负无穷表示失败\n", " 'error_message': result_entry['error']\n", " }\n", " )\n", "\n", " builtins.print = origin_print\n", " print(\"\\n--- 网格搜索回测完毕 ---\")\n", "\n", " # --- 5. 后处理和最佳结果选择 ---\n", " if all_results:\n", " results_df = pd.DataFrame(all_results)\n", " # print(\"\\n--- 所有回测结果汇总 ---\")\n", " # # 确保打印时浮点数格式化\n", " # pd.set_option('display.float_format', lambda x: '%.4f' % x)\n", " # print(results_df.to_string())\n", "\n", " # 找到最佳组合 (排除有错误的)\n", " # 过滤掉包含 'error' 键的行，或者 'error' 键的值不为空的行\n", " # 同时确保优化指标是数值，并且不为无穷大\n", " print(results_df.info())\n", " successful_results_df = results_df[(pd.to_numeric(results_df[optimization_metric], errors='coerce').notna()) &\n", " (pd.to_numeric(results_df[optimization_metric], errors='coerce') != float('-inf'))\n", " ].copy() # 使用 .copy() 避免 SettingWithCopyWarning\n", " \n", " if not successful_results_df.empty and optimization_metric in successful_results_df.columns:\n", " # 确保优化指标列是数值类型\n", " successful_results_df[optimization_metric] = pd.to_numeric(successful_results_df[optimization_metric], errors='coerce')\n", "\n", " if not successful_results_df.empty and optimization_metric in successful_results_df.columns:\n", " # 过滤掉NaN值，如果所有夏普比率都是NaN，则可能没有有效结果\n", " normal_results = successful_results_df[(results_df['total_trades'] > 200) & (results_df['total_return'] > 0)]\n", " if len(normal_results) > 0:\n", " best_result = normal_results.loc[(normal_results[optimization_metric].idxmax())]\n", " print(\"\\n--- 最优参数组合 (按夏普比率) ---\")\n", " print(best_result)\n", " else:\n", " print('ERROR!!!!!!!!!!!!!!!!!!!!')\n", " \n", " # 找到最大值的索引\n", " # best_result = successful_results_df.loc[successful_results_df[optimization_metric].idxmax()]\n", " # print(f\"\\n--- 最优参数组合 (按 {optimization_metric}) ---\")\n", " # print(best_result)\n", "\n", " # 导出到CSV\n", " output_filename = f\"grid_search_results_{datetime.now().strftime('%Y%m%d_%H%M%S')}.csv\"\n", " # results_df.to_csv(output_filename, index=False, encoding='utf-8')\n", " # print(f\"\\n所有结果已导出到: {output_filename}\")\n", "\n", " # 打印枢轴表\n", " grid_df = pd.DataFrame(grid_results)\n", " # 确保优化指标列是数值类型，非数值的（如 -inf）在pandas中可能被正确处理\n", " grid_df[optimization_metric] = pd.to_numeric(grid_df[optimization_metric], errors='coerce')\n", "\n", " pivot_table = grid_df.pivot_table(\n", " index=param1_name, columns=param2_name, values=optimization_metric\n", " )\n", " print(f\"\\n{optimization_metric} 网格结果 (Pivoted):\")\n", " print(pivot_table.to_string())\n", " else:\n", " print(f\"\\n没有成功的组合结果可供分析，或优化指标 '{optimization_metric}' 不在结果中，或所有组合均失败。\")\n", " else:\n", " print(\"没有可用的回测结果。\")\n", " print(\"\\n--- 动态网格搜索完成 ---\")\n", "\n", " # --- 6. 可视化 (依赖 GridSearchAnalyzer) ---\n", " if grid_results:\n", " grid_analyzer = GridSearchAnalyzer(grid_results, optimization_metric)\n", " grid_analyzer.find_best_parameters() # 这会找到并打印最佳参数\n", " grid_analyzer.plot_heatmap() # 这会绘制热力图\n", " else:\n", " print(\"\\n没有生成任何网格搜索结果，无法进行分析。\")" ] } ], "metadata": { "kernelspec": { "display_name": "quant", "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": "3.12.11" } }, "nbformat": 4, "nbformat_minor": 5 }