diff --git a/tests/backtesting/genetic_algorithm.ipynb b/tests/backtesting/genetic_algorithm.ipynb new file mode 100644 index 00000000..c9e92e12 --- /dev/null +++ b/tests/backtesting/genetic_algorithm.ipynb @@ -0,0 +1,189 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import random\n", + "import multiprocessing\n", + "import numpy as np\n", + "from deap import creator, base, tools, algorithms\n", + "from vnpy.app.cta_strategy.backtesting import BacktestingEngine\n", + "from boll_channel_strategy import BollChannelStrategy\n", + "from datetime import datetime\n", + "import multiprocessing #多进程\n", + "from scoop import futures #多进程" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def parameter_generate():\n", + " '''\n", + " 根据设置的起始值，终止值和步进，随机生成待优化的策略参数\n", + " '''\n", + " parameter_list = []\n", + " p1 = random.randrange(4,50,2) #布林带窗口\n", + " p2 = random.randrange(4,50,2) #布林带通道阈值\n", + " p3 = random.randrange(4,50,2) #CCI窗口\n", + " p4 = random.randrange(18,40,2) #ATR窗口 \n", + "\n", + " parameter_list.append(p1)\n", + " parameter_list.append(p2)\n", + " parameter_list.append(p3)\n", + " parameter_list.append(p4)\n", + "\n", + " return parameter_list" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def object_func(strategy_avg):\n", + " \"\"\"\n", + " 本函数为优化目标函数，根据随机生成的策略参数，运行回测后自动返回2个结果指标：收益回撤比和夏普比率\n", + " \"\"\"\n", + " # 创建回测引擎对象\n", + " engine = BacktestingEngine()\n", + " engine.set_parameters(\n", + " vt_symbol=\"IF88.CFFEX\",\n", + " interval=\"1m\",\n", + " start=datetime(2018, 9, 1),\n", + " end=datetime(2019, 1,1),\n", + " rate=0,\n", + " slippage=0,\n", + " size=300,\n", + " pricetick=0.2,\n", + " capital=1_000_000,\n", + " )\n", + "\n", + " setting = {'boll_window': strategy_avg[0], #布林带窗口\n", + " 'boll_dev': strategy_avg[1], #布林带通道阈值\n", + " 'cci_window': strategy_avg[2], #CCI窗口\n", + " 'atr_window': strategy_avg[3],} #ATR窗口 \n", + "\n", + " #加载策略 \n", + " #engine.initStrategy(TurtleTradingStrategy, setting)\n", + " engine.add_strategy(BollChannelStrategy, setting)\n", + " engine.load_data()\n", + " engine.run_backtesting()\n", + " engine.calculate_result()\n", + " result = engine.calculate_statistics(Output=False)\n", + "\n", + " return_drawdown_ratio = round(result['return_drawdown_ratio'],2) #收益回撤比\n", + " sharpe_ratio= round(result['sharpe_ratio'],2) #夏普比率\n", + " return return_drawdown_ratio , sharpe_ratio" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "\n", + "#设置优化方向：最大化收益回撤比，最大化夏普比率\n", + "creator.create(\"FitnessMulti\", base.Fitness, weights=(1.0, 1.0)) # 1.0 求最大值；-1.0 求最小值\n", + "creator.create(\"Individual\", list, fitness=creator.FitnessMulti)\n", + "\n", + "def optimize():\n", + " \"\"\"\"\"\" \n", + " toolbox = base.Toolbox() #Toolbox是deap库内置的工具箱，里面包含遗传算法中所用到的各种函数\n", + "\n", + " # 初始化 \n", + " toolbox.register(\"individual\", tools.initIterate, creator.Individual,parameter_generate) # 注册个体：随机生成的策略参数parameter_generate() \n", + " toolbox.register(\"population\", tools.initRepeat, list, toolbox.individual) #注册种群：个体形成种群 \n", + " toolbox.register(\"mate\", tools.cxTwoPoint) #注册交叉：两点交叉 \n", + " toolbox.register(\"mutate\", tools.mutUniformInt,low = 4,up = 40,indpb=0.6) #注册变异：随机生成一定区间内的整数\n", + " toolbox.register(\"evaluate\", object_func) #注册评估：优化目标函数object_func() \n", + " toolbox.register(\"select\", tools.selNSGA2) #注册选择:NSGA-II(带精英策略的非支配排序的遗传算法)\n", + " #pool = multiprocessing.Pool()\n", + " #toolbox.register(\"map\", pool.map)\n", + " #toolbox.register(\"map\", futures.map)\n", + "\n", + " #遗传算法参数设置\n", + " MU = 40 #设置每一代选择的个体数\n", + " LAMBDA = 160 #设置每一代产生的子女数\n", + " pop = toolbox.population(400) #设置族群里面的个体数量\n", + " CXPB, MUTPB, NGEN = 0.5, 0.35,40 #分别为种群内部个体的交叉概率、变异概率、产生种群代数\n", + " hof = tools.ParetoFront() #解的集合：帕累托前沿(非占优最优集)\n", + "\n", + " #解的集合的描述统计信息\n", + " #集合内平均值，标准差，最小值，最大值可以体现集合的收敛程度\n", + " #收敛程度低可以增加算法的迭代次数\n", + " stats = tools.Statistics(lambda ind: ind.fitness.values)\n", + " np.set_printoptions(suppress=True) #对numpy默认输出的科学计数法转换\n", + " stats.register(\"mean\", np.mean, axis=0) #统计目标优化函数结果的平均值\n", + " stats.register(\"std\", np.std, axis=0) #统计目标优化函数结果的标准差\n", + " stats.register(\"min\", np.min, axis=0) #统计目标优化函数结果的最小值\n", + " stats.register(\"max\", np.max, axis=0) #统计目标优化函数结果的最大值\n", + "\n", + " #运行算法\n", + " algorithms.eaMuPlusLambda(pop, toolbox, MU, LAMBDA, CXPB, MUTPB, NGEN, stats,\n", + " halloffame=hof) #esMuPlusLambda是一种基于(μ+λ)选择策略的多目标优化分段遗传算法\n", + "\n", + " return pop" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": { + "scrolled": true + }, + "outputs": [], + "source": [ + "optimize()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "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": "3.7.1" + }, + "toc": { + "base_numbering": 1, + "nav_menu": {}, + "number_sections": true, + "sideBar": true, + "skip_h1_title": false, + "title_cell": "Table of Contents", + "title_sidebar": "Contents", + "toc_cell": false, + "toc_position": {}, + "toc_section_display": true, + "toc_window_display": false + } + }, + "nbformat": 4, + "nbformat_minor": 2 +}