1. # coding=utf-8
      2. from __future__ import print_function, absolute_import, unicode_literals
      3. from datetime import datetime
      4. import numpy as np
      5. from gm.api import *
      6. import sys
      7. try:
      8.     from sklearn import svm
      9. except:
      10.     print('请安装scikit-learn库和带mkl的numpy')
      11.     sys.exit(-1)
      13. '''
      14. 本策略选取了七个特征变量组成了滑动窗口长度为15天的训练集,随后训练了一个二分类(上涨/下跌)的支持向量机模型.
      15. 若没有仓位则在每个星期一的时候输入标的股票近15个交易日的特征变量进行预测,并在预测结果为上涨的时候购买标的.
      16. 若已经持有仓位则在盈利大于10%的时候止盈,在星期五损失大于2%的时候止损.
      17. 特征变量为:1.收盘价/均值2.现量/均量3.最高价/均价4.最低价/均价5.现量6.区间收益率7.区间标准差
      18. 训练数据为:SHSE.600000浦发银行,时间从2016-03-01到2017-06-30
      19. 回测时间为:2017-07-01 09:00:00到2017-10-01 09:00:00
      20. '''
      23. def init(context):
      24.     # 订阅浦发银行的分钟bar行情
      25.     context.symbol = 'SHSE.600000'
      26.     subscribe(symbols=context.symbol, frequency='60s')
      27.     start_date = '2016-03-01'  # SVM训练起始时间
      28.     end_date = '2017-06-30'  # SVM训练终止时间
      29.     # 用于记录工作日
      30.     # 获取目标股票的daily历史行情
      31.     recent_data = history(context.symbol, frequency='1d', start_time=start_date, end_time=end_date, fill_missing='last',
      32.                           df=True)
      33.     days_value = recent_data['bob'].values
      34.     days_close = recent_data['close'].values
      35.     days = []
      36.     # 获取行情日期列表
      37.     print('准备数据训练SVM')
      38.     for i in range(len(days_value)):
      39.         days.append(str(days_value[i])[0:10])
      41.     x_all = []
      42.     y_all = []
      43.     for index in range(15, (len(days) - 5)):
      44.         # 计算三星期共15个交易日相关数据
      45.         start_day = days[index - 15]
      46.         end_day = days[index]
      47.         data = history(context.symbol, frequency='1d', start_time=start_day, end_time=end_day, fill_missing='last',
      48.                        df=True)
      49.         close = data['close'].values
      50.         max_x = data['high'].values
      51.         min_n = data['low'].values
      52.         amount = data['amount'].values
      53.         volume = []
      54.         for i in range(len(close)):
      55.             volume_temp = amount[i] / close[i]
      56.             volume.append(volume_temp)
      58.         close_mean = close[-1] / np.mean(close)  # 收盘价/均值
      59.         volume_mean = volume[-1] / np.mean(volume)  # 现量/均量
      60.         max_mean = max_x[-1] / np.mean(max_x)  # 最高价/均价
      61.         min_mean = min_n[-1] / np.mean(min_n)  # 最低价/均价
      62.         vol = volume[-1]  # 现量
      63.         return_now = close[-1] / close[0]  # 区间收益率
      64.         std = np.std(np.array(close), axis=0)  # 区间标准差
      66.         # 将计算出的指标添加到训练集X
      67.         # features用于存放因子
      68.         features = [close_mean, volume_mean, max_mean, min_mean, vol, return_now, std]
      69.         x_all.append(features)
      71.     # 准备算法需要用到的数据
      72.     for i in range(len(days_close) - 20):
      73.         if days_close[i + 20] > days_close[i + 15]:
      74.             label = 1
      75.         else:
      76.             label = 0
      77.         y_all.append(label)
      79.     x_train = x_all[: -1]
      80.     y_train = y_all[: -1]
      81.     # 训练SVM
      82.     context.clf = svm.SVC(C=1.0, kernel='rbf', degree=3, gamma='auto', coef0=0.0, shrinking=True, probability=False,
      83.                           tol=0.001, cache_size=200, verbose=False, max_iter=-1,
      84.                           decision_function_shape='ovr', random_state=None)
      85.     context.clf.fit(x_train, y_train)
      86.     print('训练完成!')
      89. def on_bar(context, bars):
      90.     bar = bars[0]
      91.     # 获取当前年月日
      92.     today = bar.bob.strftime('%Y-%m-%d')
      93.     # 获取数据并计算相应的因子
      94.     # 于星期一的09:31:00进行操作
      95.     # 当前bar的工作日
      96.     weekday = datetime.strptime(today, '%Y-%m-%d').isoweekday()
      97.     # 获取模型相关的数据
      98.     # 获取持仓
      99.     position = context.account().position(symbol=context.symbol, side=PositionSide_Long)
      100.     # 如果bar是新的星期一且没有仓位则开始预测
      101.     if not position and weekday == 1:
      102.         # 获取预测用的历史数据
      103.         data = history_n(symbol=context.symbol, frequency='1d', end_time=today, count=15,
      104.                          fill_missing='last', df=True)
      105.         close = data['close'].values
      106.         train_max_x = data['high'].values
      107.         train_min_n = data['low'].values
      108.         train_amount = data['amount'].values
      109.         volume = []
      110.         for i in range(len(close)):
      111.             volume_temp = train_amount[i] / close[i]
      112.             volume.append(volume_temp)
      114.         close_mean = close[-1] / np.mean(close)
      115.         volume_mean = volume[-1] / np.mean(volume)
      116.         max_mean = train_max_x[-1] / np.mean(train_max_x)
      117.         min_mean = train_min_n[-1] / np.mean(train_min_n)
      118.         vol = volume[-1]
      119.         return_now = close[-1] / close[0]
      120.         std = np.std(np.array(close), axis=0)
      122.         # 得到本次输入模型的因子
      123.         features = [close_mean, volume_mean, max_mean, min_mean, vol, return_now, std]
      124.         features = np.array(features).reshape(1, -1)
      125.         prediction = context.clf.predict(features)[0]
      126.         # 若预测值为上涨则开仓
      127.         if prediction == 1:
      128.             # 获取昨收盘价
      129.             context.price = close[-1]
      130.             # 把浦发银行的仓位调至95%
      131.             order_target_percent(symbol=context.symbol, percent=0.95, order_type=OrderType_Market,
      132.                                  position_side=PositionSide_Long)
      133.             print('SHSE.600000以市价单开多仓到仓位0.95')
      134.     # 当涨幅大于10%,平掉所有仓位止盈
      135.     elif position and bar.close / context.price >= 1.10:
      136.         order_close_all()
      137.         print('SHSE.600000以市价单全平多仓止盈')
      138.     # 当时间为周五并且跌幅大于2%时,平掉所有仓位止损
      139.     elif position and bar.close / context.price < 1.02 and weekday == 5:
      140.         order_close_all()
      141.         print('SHSE.600000以市价单全平多仓止损')
      144. if __name__ == '__main__':
      145.     '''
      146.     strategy_id策略ID,由系统生成
      147.     filename文件名,请与本文件名保持一致
      148.     mode实时模式:MODE_LIVE回测模式:MODE_BACKTEST
      149.     token绑定计算机的ID,可在系统设置-密钥管理中生成
      150.     backtest_start_time回测开始时间
      151.     backtest_end_time回测结束时间
      152.     backtest_adjust股票复权方式不复权:ADJUST_NONE前复权:ADJUST_PREV后复权:ADJUST_POST
      153.     backtest_initial_cash回测初始资金
      154.     backtest_commission_ratio回测佣金比例
      155.     backtest_slippage_ratio回测滑点比例
      156.     '''
      157.     run(strategy_id='strategy_id',
      158.         filename='main.py',
      159.         mode=MODE_BACKTEST,
      160.         token='token_id',
      161.         backtest_start_time='2017-07-01 09:00:00',
      162.         backtest_end_time='2017-10-01 09:00:00',
      163.         backtest_adjust=ADJUST_PREV,
      164.         backtest_initial_cash=10000000,
      165.         backtest_commission_ratio=0.0001,
      166.         backtest_slippage_ratio=0.0001)

    机器学习(股票)

    原文: https://www.myquant.cn/docs/python_strategyies/112