1纸淮、前置準(zhǔn)備

數(shù)據(jù)來源使用tushare pro平斩，具體操作請看鏈接，注冊就可以使用了
Tushare金融大數(shù)據(jù)開放社區(qū)

2咽块、直接放源碼

封裝沒有做好绘面，很多重復(fù)代碼，湊合看吧侈沪，python不太會揭璃，有大佬看到麻煩指教下是否有問題

# -*- coding: utf-8 -*-
import datetime
import os
import time

import numpy as np
import tushare as ts
from sklearn import svm
import joblib


class SvmUtil(object):

    def __init__(self):
        self.pro = ts.pro_api('此處放tushare的token')

    def svm_learning(self, stockCode):
        end_time = time.strftime('%Y%m%d', time.localtime(time.time()))
        start_year = int(time.strftime('%Y', time.localtime(time.time()))) - 2
        month_day = time.strftime('%m%d', time.localtime(time.time()))
        start_time = '{}{}'.format(start_year, month_day)
        # 獲取數(shù)據(jù)
        df = self.pro.daily(ts_code=stockCode, start_date=start_time, end_date=end_time)

        days_value = df['trade_date'].values[::-1]
        days_close = df['close'].values[::-1]
        days = []
        # 獲取行情日期列表
        for i in range(len(days_value)):
            days.append(str(days_value[i]))

        x_all = []
        y_all = []
        for index in range(15, (len(days) - 5)):
            # 計算三星期共15個交易日相關(guān)數(shù)據(jù)
            start_day = days[index - 15]
            end_day = days[index]
            data = self.pro.daily(ts_code=stockCode, start_date=start_day, end_date=end_day)
            open = data['open'].values[::-1]
            close = data['close'].values[::-1]
            max_x = data['high'].values[::-1]
            min_n = data['low'].values[::-1]
            amount = data['amount'].values[::-1]
            volume = []
            for i in range(len(close)):
                volume_temp = amount[i] / close[i]
                volume.append(volume_temp)

            open_mean = open[-1] / np.mean(open)  # 開盤價/均值
            close_mean = close[-1] / np.mean(close)  # 收盤價/均值
            diff_close_open_mean = close_mean - open_mean  # 收盤價均值-開盤價均值
            volume_mean = volume[-1] / np.mean(volume)  # 現(xiàn)量/均量
            max_mean = max_x[-1] / np.mean(max_x)  # 最高價/均價
            min_mean = min_n[-1] / np.mean(min_n)  # 最低價/均價
            diff_max_min_mean = max_mean - min_mean  # 最高價均值-最低價均值
            vol = volume[-1]
            return_now = close[-1] / close[0]  # 區(qū)間收益率
            std = np.std(np.array(close), axis=0)  # 區(qū)間標(biāo)準(zhǔn)差

            # 將計算出的指標(biāo)添加到訓(xùn)練集X
            # features用于存放因子
            # features = [close_mean, volume_mean, max_mean, min_mean, vol, return_now, std]
            features = [open_mean, close_mean, diff_close_open_mean, volume_mean, max_mean, min_mean, diff_max_min_mean,
                        vol, return_now, std]
            x_all.append(features)

        # 準(zhǔn)備算法需要用到的數(shù)據(jù)
        for i in range(len(days_close) - 20):
            if days_close[i + 20] > days_close[i + 15]:
                label = 1
            else:
                label = 0
            y_all.append(label)

        x_train = x_all[: -1]
        y_train = y_all[: -1]
        # 訓(xùn)練SVM
        model = svm.SVC(C=1.0, kernel='rbf', degree=3, gamma='auto', coef0=0.0, shrinking=True, probability=False,
                        tol=0.001, cache_size=400, verbose=False, max_iter=-1,
                        decision_function_shape='ovr', random_state=None)
        model.fit(x_train, y_train)
        joblib.dump(model, stockCode[:-3] + "_model.m")

    def svm_predict(self, stockCode):
        if not (os.path.exists(stockCode[:-3] + "_model.m")):
            self.svm_learning(stockCode)
        today = datetime.date.today()
        first = today.replace(day=1)
        last_month = first - datetime.timedelta(days=15)
        start_time = last_month.strftime("%Y%m%d")
        end_time = time.strftime('%Y%m%d', time.localtime(time.time()))
        model = joblib.load(stockCode[:-3] + "_model.m")
        df = self.pro.daily(ts_code=stockCode, start_date=start_time, end_date=end_time)
        open = df['open'].values[::-1]
        close = df['close'].values[::-1]
        train_max_x = df['high'].values[::-1]
        train_min_n = df['low'].values[::-1]
        train_amount = df['amount'].values[::-1]
        volume = []
        for i in range(len(close)):
            volume_temp = train_amount[i] / close[i]
            volume.append(volume_temp)

        open_mean = open[-1] / np.mean(open)
        close_mean = close[-1] / np.mean(close)
        diff_close_open_mean = close_mean - open_mean
        volume_mean = volume[-1] / np.mean(volume)
        max_mean = train_max_x[-1] / np.mean(train_max_x)
        min_mean = train_min_n[-1] / np.mean(train_min_n)
        diff_max_min_mean = max_mean - min_mean
        vol = volume[-1]
        return_now = close[-1] / close[0]
        std = np.std(np.array(close), axis=0)

        # 得到本次輸入模型的因子
        # features = [close_mean, volume_mean, max_mean, min_mean, vol, return_now, std]
        features = [open_mean, close_mean, diff_close_open_mean, volume_mean, max_mean, min_mean, diff_max_min_mean,
                    vol, return_now, std]
        features = np.array(features).reshape(1, -1)
        prediction = model.predict(features)[0]
        return prediction


if __name__ == '__main__':
    code = '002277.SZ'
    # SvmUtil().svm_learning(code)
    SvmUtil().svm_predict(code)

本文由博客一文多發(fā)平臺 OpenWrite 發(fā)布！