聚宽源码40 | 易学教程

原文策略源码如下：
#无模型评估的多因子模型
import pandas as pd
import numpy as np
import math
from sklearn.svm import SVR
from sklearn.model_selection import GridSearchCV
from sklearn.model_selection import learning_curve
from sklearn.linear_model import LinearRegression
from sklearn.ensemble import RandomForestRegressor
import jqdata

#初始化函数
def initialize(context):
set_params()
set_backtest()
run_daily(trade, ‘every_bar’)

#设置参数函数，调仓
def set_params():
g.days = 0
g.refresh_rate = 10
g.stocknum = 10

#测试函数
def set_backtest():
set_benchmark(‘000001.XSHG’)#基准指数
set_option(‘use_real_price’, True)
log.set_level(‘order’, ‘error’)

#设置交易函数
def trade(context):
if g.days % 10 == 0: #若持有10天，进行调仓
sample = get_index_stocks(‘000001.XSHG’, date = None) #获得股票池
q = query(valuation.code, valuation.market_cap, balance.total_assets - balance.total_liability,
balance.total_assets / balance.total_liability, income.net_profit, income.net_profit + 1,
indicator.inc_revenue_year_on_year, balance.development_expenditure).filter(valuation.code.in_(sample))
df = get_fundamentals(q, date = None)
df.columns = [‘code’, ‘log_mcap’, ‘log_NC’, ‘LEV’, ‘NI_p’, ‘NI_n’, ‘g’, ‘log_RD’] #回归的各因子

    #对因子取log
    df['log_mcap'] = np.log(df['log_mcap'])
    df['log_NC'] = np.log(df['log_NC'])
    df['NI_p'] = np.log(np.abs(df['NI_p']))
    df['NI_n'] = np.log(np.abs(df['NI_n'][df['NI_n']<0]))
    df['log_RD'] = np.log(df['log_RD'])
    df.index = df.code.values
    del df['code']
    df = df.fillna(0)
    df[df>10000] = 10000
    df[df<-10000] = -10000
    industry_set = ['801010', '801020', '801030', '801040', '801050', '801080', '801110', '801120', '801130', 
              '801140', '801150', '801160', '801170', '801180', '801200', '801210', '801230', '801710',
              '801720', '801730', '801740', '801750', '801760', '801770', '801780', '801790', '801880','801890'] #申万一级指数
    
    for i in range(len(industry_set)):
        industry = get_industry_stocks(industry_set[i], date = None)
        s = pd.Series([0]*len(df), index=df.index)
        s[set(industry) & set(df.index)]=1
        df[industry_set[i]] = s
        
        #回归X,Y
    X = df[['log_NC', 'LEV', 'NI_p', 'NI_n', 'g', 'log_RD','801010', '801020', '801030', '801040', '801050', 
            '801080', '801110', '801120', '801130', '801140', '801150', '801160', '801170', '801180', '801200', 
            '801210', '801230', '801710', '801720', '801730', '801740', '801750', '801760', '801770', '801780', 
            '801790', '801880', '801890']]
    Y = df[['log_mcap']]
    X = X.fillna(0)
    Y = Y.fillna(0)
    
    #支持向量回归（Support Vector Regression)
    svr = SVR(kernel='rbf', gamma=0.1) 
    model = svr.fit(X, Y)  #将XY带入SVR
    factor = Y - pd.DataFrame(svr.predict(X), index = Y.index, columns = ['log_mcap']) #获取残差值
    factor = factor.sort_index(by = 'log_mcap')
    stockset = list(factor.index[:10])     #按残差大小分成10组
    sell_list = list(context.portfolio.positions.keys())
    for stock in sell_list:
        if stock not in stockset[:g.stocknum]:    #如果股票不在该档就卖出
            stock_sell = stock
            order_target_value(stock_sell, 0)
        
    if len(context.portfolio.positions) < g.stocknum:  #如果持仓天数小于10，持仓期平均cash剩余
        num = g.stocknum - len(context.portfolio.positions)
        cash = context.portfolio.cash/num
    else:
        cash = 0
        num = 0
    for stock in stockset[:g.stocknum]:   #如果持仓股票仍在该档，则不卖出
        if stock in sell_list:
            pass
        else:
            stock_buy = stock
            order_target_value(stock_buy, cash)
            num = num - 1
            if num == 0:
                break
    g.days += 1                            #按天进行
else:
    g.days = g.days + 1

来源：CSDN

作者：程序化小学生

链接：https://blog.csdn.net/weixin_45131220/article/details/103489597

标签

源码

log