features_lstm.py

from Klang import (Kl,Klang_init,
    C,O,V,H,L, CLOSE,HIGH,DATETIME,
    MA,CROSS,BARSLAST,HHV,LLV,COUNT,BARSLASTFIND,
    MAX,MIN,MACD)
from Klang.common import end as today
import talib 

import sys
import linecache
import pandas as pd
import requests,time
import json

def PrintException():
    exc_type, exc_obj, tb = sys.exc_info()
    f = tb.tb_frame
    lineno = tb.tb_lineno
    filename = f.f_code.co_filename
    linecache.checkcache(filename)
    line = linecache.getline(filename, lineno, f.f_globals)
    print ('EXCEPTION IN ({}, LINE {} "{}"): {}'.format(filename, lineno, line.strip(), exc_obj))

all_dict = {}
target_day = 10 #收盘价之后的几天内最高价格，判断是否有涨价空间
hostname = "http://klang.org.cn"
hostname = "http://klang.zhanluejia.net.cn"


def save_dict_tofile(datas,filename):
    content = json.dumps(datas)
    f = open(filename,"w+")
    f.write(content)
    f.close() 

def get_features(code,end):
    try:
        json = requests.get(hostname+"/features",
            params={"code":code,"end":end,"limit":200},timeout=1000).json()
    except:
        time.sleep(2)
        json = requests.get(hostname+"/features",
            params={"code":code,"end":end,"limit":200},timeout=1000).json()

    df = pd.json_normalize(json)
    if len(df) < 1:
       return []
    df = df.drop(columns=['_id','codedate','id'])
    datas = df.sort_values(by="date",ascending=True) 

    return datas

def main_loop(start,endday):
    global all_dict

    #for df in Kl.df_all[:100]:
    for df in Kl.df_all:


        Kl.code(df["code"])

        if start is None:
            Kl.date(end=endday)
        else:
            Kl.date(start=start,end=endday)
        try:
            if len(Kl.currentdf['df']) <= target_day:
                continue

            allDate = DATETIME.data
            # 如果target_day = N,表示，最后的N 天数据不能作为训练或者测试数据
            # 我们会计算这个 N 天的最大值作为目标值，计算涨幅空间

            featureday = allDate[-target_day]
            datas = get_features(df['code'],featureday)   
            print(df['code'],df['name'])
            datas = datas[(datas['date'] >= allDate[0]) & (datas['date'] < featureday)]

            max_target = talib.MAX(C.data,target_day)
            rise_target = (max_target[target_day:].values / C.data[:-target_day].values - 1 ) * 100

            datas['oc'] = (O.data[:-target_day].values / C.data[:-target_day].values - 1)*100
            datas['close'] = (C.data[:-target_day] / C.data[0]).values

            datas['target'] = rise_target 
            all_dict[df['code']] = datas.to_json()
        except KeyboardInterrupt:
            break
        except:
            PrintException()


fields = [
       'code', 'date', 'dea', 'diff', 'ma10', 'ma120', 'ma20', 'ma30', 'ma5',
       'ma60', 'macd', 'name', 'rise', 'risevol','oc','close',
       'target']

Klang_init()

main_loop(start=None,endday='2021-12-01')

save_dict_tofile(all_dict,'lstm_train'+today+'.csv')