Skip to content
TheChatArchiveThe Chat Archive
gpt-4-0314· OpenAIAi2 WildChat-4.8M (AI2)

def getkdj(stockcode, enddate): Step1: 获取数据 startdate = '2022-11-01' lg = bs.login() rs…

  1. User

    def getkdj(stockcode, end_date): # Step1: 获取数据 start_date = '2022-11-01' lg = bs.login() rs = bs.query_history_k_data_plus(code, "date,code,open,high,low,close,volume", start_date=start_date, end_date=end_date, frequency="d", adjustflag='2') # 注意adjustflag取2为前复权 data_list = [] while (rs.error_code == '0') & rs.next(): data_list.append(rs.get_row_data()) df = pd.DataFrame(data_list, columns=rs.fields) df[['open', 'high', 'low', 'close', 'volume']] = df[['open', 'high', 'low', 'close', 'volume']].astype( 'float64') df = df.rename(columns={'date': 'datetime'}) df.index = pd.DatetimeIndex(df['datetime']) bs.logout()

    lowList = df['low'].rolling(19).min()  # 计算low值19日移动最低
    lowList.fillna(value=df['low'].expanding().min(), inplace=True)
    highList = df['high'].rolling(19).max()  # 计算high值19日移动最高
    highList.fillna(value=df['high'].expanding().max(), inplace=True)
    rsv = (df.loc[:, 'close'] - lowList) / (highList - lowList) * 100
    df.loc[:, 'kdj_k'] = rsv.ewm(com=6).mean()
    df.loc[:, 'kdj_d'] = df.loc[:, 'kdj_k'].ewm(com=2).mean()
    df.loc[:, 'kdj_j'] = 3.0 * df.loc[:, 'kdj_k'] - 2.0 * df.loc[:, 'kdj_d']
    df = df.sort_index(ascending=False)
    print(df)
    return df['kdj_j'][len(df)-1]
    

    这段代码用python内建方法替换pandas,如何改写,要求不要用到第三方库

  2. Assistant

    你可以使用Python内建的数据结构和方法重写这个代码,但需要注意,这将使代码变得更加复杂和难以维护。以下是一个可能的实现:

    import datetime
    import time
    
    def getkdj(stockcode, end_date):
        # Step1: 获取数据
        start_date = '2022-11-01'
        lg = bs.login()
        rs = bs.query_history_k_data_plus(code,
                                          "date,code,open,high,low,close,volume",
                                          start_date=start_date, end_date=end_date, frequency="d",
                                          adjustflag='2')  # 注意adjustflag取2为前复权
        data_list = []
        while (rs.error_code == '0') & rs.next():
            data_list.append(rs.get_row_data())
        
        index = ['datetime', 'code', 'open', 'high', 'low', 'close', 'volume']
        df_dict = {key: [] for key in index}
        for row in data_list:
            for i in range(len(index)):
                df_dict[index[i]].append(float(row[i]) if i > 1 else row[i])
        
        def rolling_min(data, window):
            result = [float('inf') for _ in range(window-1)] + [min(data[i:i+window]) for i in range(len(data) - window + 1)]
            return result
        
        def rolling_max(data, window):
            result = [float('-inf') for _ in range(window-1)] + [max(data[i:i+window]) for i in range(len(data) - window + 1)]
            return result
        
        lowList = rolling_min(df_dict['low'], 19)
        highList = rolling_max(df_dict['high'], 19)
        rsv = [(close - low) / (high - low) * 100 for close, low, high in zip(df_dict['close'], lowList, highList)]
        
        df_dict['kdj_k'] = []
        df_dict['kdj_d'] = []
        df_dict['kdj_j'] = []
        
        k, d = rsv[0], rsv[0]
        for rsv_val in rsv:
            k = k * (1 - 1 / (6 + 1)) + rsv_val * 1 / (6 + 1)
            d = d * (1 - 1 / (2 + 1)) + k * 1 / (2 + 1)
            j = 3 * k - 2 * d
            df_dict['kdj_k'].append(k)
            df_dict['kdj_d'].append(d)
            df_dict['kdj_j'].append(j)
        
        d = datetime.datetime.strptime(end_date,'%Y-%m-%d')
        end_timestamp = time.mktime(d.timetuple())
        last_kdj_j = None
        
        for dt, jdj in zip(reversed(df_dict['datetime']), reversed(df_dict['kdj_j'])):
            cur_d = datetime.datetime.strptime(dt,'%Y-%m-%d')
            cur_timestamp = time.mktime(cur_d.timetuple())
            if cur_timestamp >= end_timestamp:
                last_kdj_j = jdj
            else:
                break
        
        print(df_dict)
        return last_kdj_j
    

    这个实现使用了原生的Python列表和字典,以及一些辅助功能,如rolling_min和rolling_max。请注意,这个实现很可能效率较低,不易于调试和维护。在实际应用中,建议还是使用Pandas等第三方库进行数据处理。

  3. User

    上面运行结果:kdj_k = nan kdj_d = nan kdj_j=nan

  4. Assistant

    我注意到一个可能的问题:在计算rsv时,如果high和low相等(例如,当数据有缺失值或极端情况时),除法将导致除以零错误,从而产生NaN结果。你可以通过添加一个小的正值来避免除以零错误:

    eps = 1e-8
    rsv = [(close - low) / (high - low + eps) * 100 for close, low, high in zip(df_dict['close'], lowList, highList)]
    

    此外,请检查从API获取的原始数据是否存在缺失值或异常值,这可能导致计算结果出现NaN。如果可能,请考虑使用数据清洗方法填充缺失值或剔除异常值。