炒股利器！用Python大模型筛选低估值股票 原创

import os
import pandas as pd
import yfinance as yf
from datetime import datetime, timedelta
from cohere_interface import Cohere  # 确保导入Cohere类
from concurrent.futures import ThreadPoolExecutor

class Screener:
    def __init__(self, folder):
        self.index = folder  # 设置索引为文件夹名称
        self.folder = f'/data/historical/analysis/{folder}/'  # 设置数据文件夹路径
        self.files = os.listdir(self.folder)  # 获取文件夹中的所有文件
        self.df_all = pd.DataFrame()  # 初始化空的数据框，用于存储最终数据

        # 创建数据文件夹
        if not os.path.exists(self.folder):
            os.makedirs(self.folder)

def process_file(self, filename):
        if (filename.endswith('xlsx') or filename.endswith('csv')) and not filename.startswith('~$'):
            file_path = os.path.join(self.folder, filename)
            print(f"Processing file: {file_path}")

            try:
                df = pd.read_csv(file_path)  # 读取CSV文件
                df['Date'] = pd.to_datetime(df['Date'], utc=True)  # 转换日期列为日期时间格式
                df['Date'] = df['Date'].dt.date  # 提取日期部分
                current_date = df.iloc[-1]['Date']  # 获取最新日期
                symbol = filename.split()[0]  # 获取股票符号
                info = yf.Ticker(symbol).info  # 获取股票信息

                # 提取关键财务数据
                pe = info.get('trailingPE', None)
                pb = info.get('priceToBook', None)
                debtToEquity = round(info.get('debtToEquity', 0) / 100.0, 2)
                dividend_yield = info.get('dividendYield', None)
                eps_growth = info.get('earningsQuarterlyGrowth', None)
                close = df.iloc[-1]['Close']

                # 生成数据字典
                data = {
                    'date': [current_date],
                    'symbol': [symbol],
                    'pe': [pe],
                    'pb': [pb],
                    'debtToEquity': [debtToEquity],
                    'dividend_yield': [dividend_yield],
                    'eps_growth': [eps_growth],
                    'close': [close]
                }

                df_change = pd.DataFrame(data)  # 将数据字典转换为数据框
                self.df_all = pd.concat([self.df_all], axis=0)  # 将新的数据框添加到总数据框中
            except Exception as e:
                print(f"Error processing file {file_path}: {e}")

def run(self):
        with ThreadPoolExecutor(max_workers=4) as executor:  # 使用4个线程并行处理
            executor.map(self.process_file, self.files)

def store_in_csv(self):
        # 将最终数据存储到CSV文件中
        folder = f'./data/historical/screener/'
        if not os.path.exists(folder):
            os.makedirs(folder)
        filename = f'{self.index}_{str(datetime.today()}.csv'
        path = os.path.join(folder, filename)
        self.df_all.to_csv(path, index=False)

def ai_screening(self):
        # 使用Cohere进行AI筛选
        co = Cohere()
        message = f"""你将获得一些股票的数据。请分析这些数据并筛选出2只被低估且未来有可能产生更高回报的股票。以下是数据 {self.df_all}"""
        preamble = """你是一名优秀的股市分析师，能够很好地理解提供的数据并做出谨慎的决策。输出应严格采用以下JSON格式：
        [
            {
                "stock": "股票名称",
                "justification": "筛选或选择该股票的理由"
            },
            {
                "stock": "股票名称",
                "justification": "筛选或选择该股票的理由"
            }
        ]"""
        result = co.chat_cohere(message=message, preamble=preamble)
        print(result)