1.1、财务数据——总市值和PE

from jqdatasdk import *
import seaborn as sns
import math
import datetime
from tqdm import tqdm
import pandas as pd
import matplotlib.pyplot as plt
import numpy as np
df = get_fundamentals(query(valuation,indicator)) 
raw_data = [] 
for index in range(len(df['code'])): 
    raw_data_item = { 
    'code' :df['code'][index], 
    'market_cap':df['market_cap'][index], 
    'pe_ratio' :df['pe_ratio'][index] 
    } 
    raw_data.append(raw_data_item)

# # 按照财务信息中的总市值降序排序
# raw_data = sorted(raw_data,key = lambda item:item['market_cap'], reverse=True) 
# # 剔除总市值排名最小的10%的股票
# fitered_market_cap = raw_data[:int(len(raw_data) * 0.9)] 
# # 剔除PE TTM 小于0或大于100
# filtered_pe = [] 
# for stock in fitered_market_cap: 
#     if stock['pe_ratio'] == None or math.isnan(stock['pe_ratio']) or float(stock['pe_ratio']) < 0 or float(stock['pe_ratio']) > 100: 
#         continue 
#     filtered_pe.append(stock['code'])

#PE数据集
df_pe=pd.DataFrame(raw_data)

df_name=get_all_securities().reset_index().rename(columns={'index':'code'})

df_pe=pd.merge(df_pe,df_name,on='code',how='left')

df_pe=df_pe[['code','market_cap','pe_ratio','display_name']]

1.2、行情数据——25日涨跌幅

current_dt=datetime.datetime.now().strftime('%Y-%m-%d')
CHANGE_PCT_DAY_NUMBER=25

def load_change_pct_data(current_dt,codes):
    change_pct_dict_list = []
    # 计算涨跌幅需要用到前一日收盘价，所以需要多加载一天的数据，
    # 而这里在第二日的开盘前运行，计算前一个交易日收盘后的数据，所以需要再多加载一天
    # 使用固定的25个交易日，而非25个bar计算涨跌幅
    count = CHANGE_PCT_DAY_NUMBER + 1
    # 获取25个交易日的日期
    pre_25_dates = get_trade_days(start_date=None, end_date=current_dt, count=count)
    pre_25_date = pre_25_dates[0]
    pre_1_date = pre_25_dates[-1]
    for code in codes:
        pre_25_data =  get_price(code, start_date=None, end_date=pre_25_date, frequency='daily', fields=['close'], skip_paused=True, fq='post', count=1)
        pre_1_data =  get_price(code, start_date=None, end_date=pre_1_date, frequency='daily', fields=['close'], skip_paused=True, fq='post', count=1)
        pre_25_close = None
        pre_1_close = None
        if str(pre_25_date) == str(pre_25_data.index[0])[:10]:
            pre_25_close = pre_25_data['close'][0]
        if str(pre_1_date) == str(pre_1_data.index[0])[:10]:
            pre_1_close = pre_1_data['close'][0]

        if pre_25_close != None and pre_1_close != None and not math.isnan(pre_25_close) and not math.isnan(pre_1_close):
            change_pct = (pre_1_close - pre_25_close) / pre_25_close
            item = {'code':code, 'change_pct': change_pct}
            change_pct_dict_list.append(item)
    return change_pct_dict_list

def load_change_pct_data_1(current_dt,code):
    change_pct_dict_list = []
    # 计算涨跌幅需要用到前一日收盘价，所以需要多加载一天的数据，
    # 而这里在第二日的开盘前运行，计算前一个交易日收盘后的数据，所以需要再多加载一天
    # 使用固定的25个交易日，而非25个bar计算涨跌幅
    count = CHANGE_PCT_DAY_NUMBER + 1
    # 获取25个交易日的日期
    pre_25_dates = get_trade_days(start_date=None, end_date=current_dt, count=count)
    pre_25_date = pre_25_dates[0]
    pre_1_date = pre_25_dates[-1]
    pre_25_data =  get_price(code, start_date=None, end_date=pre_25_date, frequency='daily', fields=['close'], skip_paused=True, fq='post', count=1)
    pre_1_data =  get_price(code, start_date=None, end_date=pre_1_date, frequency='daily', fields=['close'], skip_paused=True, fq='post', count=1)
    pre_25_close = None
    pre_1_close = None
    if str(pre_25_date) == str(pre_25_data.index[0])[:10]:
        pre_25_close = pre_25_data['close'][0]
    if str(pre_1_date) == str(pre_1_data.index[0])[:10]:
        pre_1_close = pre_1_data['close'][0]

    if pre_25_close != None and pre_1_close != None and not math.isnan(pre_25_close) and not math.isnan(pre_1_close):
        change_pct = (pre_1_close - pre_25_close) / pre_25_close
        item = {'code':code, 'change_pct': change_pct}
        change_pct_dict_list.append(item)
    return change_pct_dict_list

# # 加载标的的涨跌幅信息
# change_pct_dict_list = load_change_pct_data(current_dt,filtered_pe) 
# # 按照涨跌幅升序排序
# change_pct_dict_list = sorted(change_pct_dict_list,key = lambda item:item['change_pct'], reverse=False) 
# # 取跌幅前10%的股票
# change_pct_dict_list = change_pct_dict_list[0:(int(len(change_pct_dict_list)*0.1))]

change_pct_dict_list=[]
for i in tqdm(df_pe['code']):
    change_pct_dict_list.append(load_change_pct_data_1(current_dt,i))

df_stocks_industry_1=pd.DataFrame(list(df_stocks_industry.items()))

df_stocks_industry_1.columns=['code','industry_name']

现在是否ST

df_stoacks_is_st=get_extras('is_st',list(df_stocks_industry_1['code']),start_date='2020-03-06', end_date='2020-03-06')

df_stoacks_is_st=df_stoacks_is_st.T.reset_index().rename(columns={'index':'code','2020-03-06 00:00:00':'is_st'})

df_stoacks_is_st.columns=['code','is_st']

特征合并

# df_pe
df_stocks_industry=pd.merge(df_pe,df_stocks_industry_1,on='code',how='left')

#25日涨跌幅 change_pct_dict_df
df_stocks_industry_change=pd.merge(df_stocks_industry,change_pct_dict_df,on='code',how='left')

#合并资产负债率特征
df_stocks_feature=pd.merge(df_stocks_industry_change,df_liability_vs_assets,on='code',how='left')

#合并流动比率特征 df_current
df_stocks_feature=pd.merge(df_stocks_feature,df_current,on='code',how='left')

#总资产周转率、存货周转率、平均在库天数
df_stocks_feature=pd.merge(df_stocks_feature,df_turnover,on='code',how='left')

#是否st标签
df_stocks_feature=pd.merge(df_stocks_feature,df_stoacks_is_st,on='code',how='left')

df_stocks_feature