Q6

2022-06-04 09:03:17 阅读：165 来源： 互联网

标签：index Q6 benchmark ret factor data stock

import pickle
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from datetime import datetime

def plot_ec(ec, ret_next='ret_next', fac='factor'):
    '''
    绘制单变量分组中各组的资金曲线
    ret_next: 分组数据中下期收益率的列名
    fac: 分组的变量的名称
    '''
    ec = ec.copy()
    ec = ec.set_index('date').dropna()
    ec['equity_curve'] = ec.groupby('group')[ret_next].apply(lambda x: (x+1).cumprod())

    group = int(ec.group.max())
    for i in range(group):
        if(i == 0):
            lab = 'low-' + fac
        elif(i == group-1):
            lab = 'high-'+ fac
        else:
            lab = 'g' + str(i+1)
        this = ec[ec.group == i+1]
        if i == 0: print('第1组最后曲线净值：',this.equity_curve[-1])
        plt.plot(this.index, this.equity_curve, label=lab)
    plt.legend()
    plt.show()

def set_group(x: pd.Series, group=10):
    '''
    获得组号, 用于groupby后的apply函数中
    group:分为几组
    '''
    size = x.size - np.count_nonzero(np.isnan(x))
    return np.ceil(x.rank(method='first') / (size/group))

def read_data():
    f1 = open('./data/CSI500_factor_data.pkl', 'rb')
    f2 = open('./data/benchmark.pkl', 'rb')
    f3 = open('./data/price_data.pkl', 'rb')

    CSI500_factor = pickle.load(f1)
    benchmark = pickle.load(f2)
    price_data = pickle.load(f3)

    return CSI500_factor, benchmark, price_data

def stock_transform(price_data):
    stock = pd.DataFrame(columns=['date', 'stkcd', 'close', 'ret_next'])
    for i in range(len(price_data.index)):
        temp = pd.DataFrame({'date':price_data.columns,
                               'stkcd': price_data.index[i],
                               'close': price_data.iloc[i, :]})
        temp['ret_next'] = temp['close'].shift(-1) / temp.close - 1
        stock = pd.concat([stock, temp], axis=0)

    return stock

def factor_merge(stock, CSI500_factor):
    stock = stock.reset_index(drop=True)
    # stock_ = stock.set_index(['stkcd', 'date'])
    stock['date'] = stock['date'].apply(lambda x: x[:6])

    CSI500_factor_ = CSI500_factor.reset_index()
    CSI500_factor_.columns = ['stkcd', 'date', 'factor_value']
    CSI500_factor_['date'] = CSI500_factor_['date'].apply(lambda x: x[:6])
    stock = pd.merge(stock, CSI500_factor_, how='left', on=['date', 'stkcd'])
    return stock


if __name__ == '__main__':
    CSI500_factor, benchmark, price_data = read_data()
    stock = stock_transform(price_data)
    stock = factor_merge(stock, CSI500_factor)
    raise TypeError('test')
    stock['group'] = stock['factor_value'].groupby(stock.date).apply(set_group, group=10)

    ec = stock.groupby(['date', 'group'])['ret_next'].agg('mean').reset_index()  # 等权重十分位投资组合下个月收益
    del stock['group']
    plot = 1
    if plot: plot_ec(ec,  fac='factor_value')

    ####################################################################################################################
    # IC & IR
    ####################################################################################################################
    IC = stock.set_index('date')[['factor_value', 'ret_next']].groupby('date').apply(lambda x: x.corr().iloc[0, 1])
    IC_value = IC.mean()
    IR = IC_value / IC.std()
    print('CSI500 factor', '\n 因子IC值：', round(IC_value, 5), '\n 因子IR值：', round(IR, 4))

    ####################################################################################################################
    # 对冲组合收益率计算
    ####################################################################################################################

    max_ret = ec.groupby(['date'])[['ret_next']].agg('max')
    benchmark.columns = ['zz500']

    benchmark_ = benchmark.copy()
    benchmark_.index = pd.to_datetime(benchmark_.index)
    benchmark_min = benchmark_.loc[benchmark_.groupby(benchmark_.index.to_period('M')).apply(lambda x: x.index.min())]
    benchmark_max = benchmark_.loc[benchmark_.groupby(benchmark_.index.to_period('M')).apply(lambda x: x.index.max())]

    index_ = np.unique([i[:6] for i in benchmark.index])
    benchmark_m = pd.DataFrame(columns=['zz_ret'], index=index_)
    benchmark_m['zz_ret'] = benchmark_max.values/ benchmark_min.values -1

    hege_ = pd.merge(max_ret, benchmark_m, left_index=True, right_index=True, how='left')
    hege_['ret'] = hege_['ret_next'] - hege_['zz_ret']
    hege_['year'] = [i[:4] for i in hege_.index]

    year_ = pd.DataFrame()
    year_['year_ret'] = hege_.groupby(['year'])['ret_next'].agg('sum')
    year_['year_std'] = hege_.groupby(['year'])['ret_next'].agg('std')
    year_['year_recall'] = hege_.groupby(['year'])['ret_next'].agg('max')

标签：index,Q6,benchmark,ret,factor,data,stock
来源： https://www.cnblogs.com/RankFan/p/Q6.html

本站声明： 1. iCode9 技术分享网（下文简称本站）提供的所有内容，仅供技术学习、探讨和分享；
2. 关于本站的所有留言、评论、转载及引用，纯属内容发起人的个人观点，与本站观点和立场无关；
3. 关于本站的所有言论和文字，纯属内容发起人的个人观点，与本站观点和立场无关；
4. 本站文章均是网友提供，不完全保证技术分享内容的完整性、准确性、时效性、风险性和版权归属；如您发现该文章侵犯了您的权益，可联系我们第一时间进行删除；
5. 本站为非盈利性的个人网站，所有内容不会用来进行牟利，也不会利用任何形式的广告来间接获益，纯粹是为了广大技术爱好者提供技术内容和技术思想的分享性交流网站。

ICode9

Q6