数据获取/多指标排名收益分布.py

import pandas as pd
import numpy as np
import choose
import os
import datetime
from pymongo import MongoClient

client = MongoClient(
    "mongodb://wth000:[email protected]:27017/dbname?authSource=wth000")
db = client["wth000"]
# names = ["可转债","COIN", "股票", "指数", "行业", "ETF",]
names = ["COIN", ]
# 获取当前.py文件的绝对路径
file_path = os.path.abspath(__file__)
# 获取当前.py文件所在目录的路径
dir_path = os.path.dirname(file_path)
# 获取当前.py文件所在目录的上两级目录的路径
dir_path = os.path.dirname(os.path.dirname(dir_path))
files = os.listdir(dir_path)
for file in files:
    for filename in names:
        if (filename in file):
            try:
                # 获取文件名和扩展名
                name, extension = os.path.splitext(file)
                df = pd.read_csv(os.path.join(dir_path, f"{name}.csv"))
                print(f'{name}')
                # df.columns = ["日期", "代码", "开盘", "收盘", "最高", "最低", "昨收", "成交额", "成交量", "振幅", "换手率", "涨跌幅", "涨跌幅（开盘）",
                #               "流通波动", "流通贡献", "资金波动", "资金贡献", "换手波动", "换手贡献", "净利润同比增长率", "净资产收益率", "归属母公司股东的净利润同比增长率", "总市值", "流通市值", "总资产", "总负债", "净资产", "净利润", "市盈率", "市净率", "市销率", "市现率", "资产负债率", "营收",
                #               "净利润（归母）TTM", "营业收入增长率", "总资产增长率", "净利润增长率", "净资产增长率"]
                # dflanchou = pd.read_csv(f"每日蓝筹.csv")
                # dflanchou = dflanchou.iloc[:, :2]
                # dflanchou.columns = ['日期', '代码']
                # # dflanchou.rename(columns={dflanchou.columns[0]: '日期', dflanchou.columns[1]: '代码'}, inplace=True)
                # df = df.merge(dflanchou[['日期', '代码']], on=['日期', '代码'])
                # print(df)
                # # df总共517万条，看看合并之后多少条
                # dfkaipan = pd.read_csv('非ST股票（分钟）开盘后五分钟.csv')
                # dfkaipan.columns = ['日期', '代码', '开盘(开盘后五分钟)', '最高(开盘后五分钟)', '最低(开盘后五分钟)',
                #                     '收盘(开盘后五分钟)', '昨收(开盘后五分钟)', '成交额(开盘后五分钟)', '涨跌幅（今收今开）(开盘后五分钟)']
                # df = df.merge(dfkaipan, on=['日期', '代码'])  # 增加设置基本面因子
                # dfshoupan = pd.read_csv('非ST股票（分钟）收盘前五分钟.csv')
                # dfshoupan.columns = ['日期', '代码', '开盘(收盘前五分钟)', '最高(收盘前五分钟)', '最低(收盘前五分钟)',
                #                           '收盘(收盘前五分钟)', '昨收(收盘前五分钟)', '成交额(收盘前五分钟)', '涨跌幅（今收今开）(收盘前五分钟)']
                # df = df.merge(dfshoupan, on=['日期', '代码'])  # 增加设置基本面因子
                print(df)
                watchtime = 1999
                # start_date = datetime.datetime(
                #     watchtime, int(1), int(1)).strftime("%Y-%m-%d %H:%M:%S")
                # end_date = datetime.datetime(datetime.datetime.strptime(
                #     start_date, "%Y-%m-%d %H:%M:%S").year + 8, int(1), int(1)).strftime("%Y-%m-%d %H:%M:%S")
                # df = df[(df["日期"] >= start_date) & (df["日期"] <= end_date)]
                df = df.groupby("代码", group_keys=False).apply(
                    choose.technology)
                df = df.groupby("日期", group_keys=False).apply(choose.rank)
                df.to_csv(f'指标（收益率隔夜）{name}.csv')
                df, m, n = choose.choose(name, df)
                if ("股票" in name):
                    for i in range(1, n+1):
                        df = df[df[f"{i}日后总涨跌幅（未来函数）"] <= 3*(1+0.1*n)]
                else:
                    for i in range(1, n+1):
                        df = df[df[f"{i}日后总涨跌幅（未来函数）"] <= 20*(1+0.1*n)]
                # 将数据划分成a个等长度的区间
                a = 50
                ranges = []
                left = 0
                right = 1
                step = (right - left) / a
                for i in range(a):
                    ranges.append((left + i * step, left + (i + 1) * step))
                # 筛选出列名中包含"rank"的列
                rank_cols = df.filter(like="rank").columns.tolist()
                # 创建空的结果DataFrame
                result_df = pd.DataFrame()
                # 循环处理每个指标和区间
                for rank_range in ranges:
                    col_result_df = pd.DataFrame()  # 创建一个空的DataFrame，用于存储指标的结果
                    for col_name in rank_cols:
                        # 根据区间筛选DataFrame
                        sub_df = df[(df[col_name] >= rank_range[0]) &
                                    (df[col_name] <= rank_range[1])]
                        # 计算收益
                        sub_df_mean = sub_df.mean(numeric_only=True)  # 均值法
                        # 构造包含指标名和涨跌幅的DataFrame，并添加到列结果DataFrame中
                        result_sub_df = pd.DataFrame(
                            {col_name: [sub_df_mean[f"{n}日后总涨跌幅（未来函数）"]]}, index=[rank_range])
                        col_result_df = pd.concat(
                            [col_result_df, result_sub_df], axis=1)
                    result_df = pd.concat([result_df, col_result_df])
                # 新建涨跌分布文件夹在上级菜单下，并保存结果
                path = os.path.join(os.path.abspath("."), "资产多指标排名收益分布")
                if not os.path.exists(path):
                    os.makedirs(path)
                result_df.to_csv(
                    f"{path}/{name}持有{n}日{str(watchtime)}年多指标排名收益分布.csv")
                print("任务已经完成！")
            except Exception as e:
                print(f"发生bug: {e}")