python_在K线找出波段_01_找出所有转折点

本文链接：https://blog.csdn.net/m0_37967652/article/details/132030255

在【PyQt5开发验证K线视觉想法工具V1.0】中查看效果

写在前面：

我们一般看K线图时通过视觉判断波段具有很强的主观性，所以通过代码找到的波段并不能满足所有人的需求。大家有自己特定波段述求的，可以自行修改代码。

寻找波段分两大步骤，本篇文章是第一步骤——寻找所有的转折点，即所有的峰值和谷值，也可以理解为所有的高点和低点。

需要考虑的几种K线图情况：

1 巨大宽幅波动

2 窄幅波动

3 极窄幅横盘

4 缓慢下跌

5 缓慢上升

PS:下面找了大致的K线图

寻找所有转折点逻辑：

1 时间区间取今年年内的日K线

2 从最近的时间往前寻找，从K线图看就是从最右端向左边开始寻找

3 从右到左，遇到凸起点记录为峰（高点），遇到凹点记录为谷（低点），直到K线走完

4 代码逻辑上，使用pandas的cummax()寻找峰，cummin()寻找谷

5 循环交替寻找，找峰后找谷，再找峰再找谷。。。，每找到第一个峰或谷就返回，下回从上一次返回的点开始寻找

代码：

寻找转折点方法：找到第一个转折点就返回

def caculate_turning_point(pre_df,start_x,y_col,x_col,mark):
    '''
    计算转折点
    :param pre_df: 要计算的数据
    :param start_x: 起点
    :param y_col: y的列名
    :param x_col: x的列名
    :param mark:True=峰 False=谷
    :return: 第一个转折点 [x_val,y_val,mark]
    '''
    df = pre_df.loc[pre_df[x_col]<start_x].copy()
    if mark:
        # 峰值
        df['p0'] = df[y_col].cummax()
        df['p1'] = df['p0'] - df['p0'].shift(1)
        df['p2'] = 1
        df.loc[df['p1'] != 0, 'p2'] = 0
        df['p3'] = 0
        df.loc[(df['p2'] == 0) & (df['p2'].shift(-1) == 1), 'p3'] = 1

        df_p = df.loc[df['p3']==1].copy()

        if len(df_p)<=0:
            return [None,None,mark]
        else:
            p_i = df_p.iloc[0][x_col]
            p_y = df_p.iloc[0][y_col]
            return [p_i,p_y,mark]
        pass
    else:
        # 谷值
        df['l0'] = df[y_col].cummin()
        df['l1'] = df['l0'] - df['l0'].shift(1)
        df['l2'] = 1
        df.loc[df['l1'] != 0, 'l2'] = 0
        df['l3'] = 0
        df.loc[(df['l2'] == 0) & (df['l2'].shift(-1) == 1), 'l3'] = 1

        df_l = df.loc[df['l3'] == 1].copy()

        if len(df_l) <= 0:
            return [None, None, mark]
        else:
            l_i = df_l.iloc[0][x_col]
            l_y = df_l.iloc[0][y_col]
            return [l_i, l_y, mark]
        pass
    pass

循环寻找峰谷方法

def circle_find(start_mark,i_start,pre_df,py_col,ly_col,x_col)->List:
    res_list = []
    i = i_start
    while True:
        if i<=0:
            break
        if start_mark:
            # 峰值
            res_one = caculate_turning_point(pre_df,i,py_col,x_col,start_mark)
        else:
            # 谷值
            res_one = caculate_turning_point(pre_df,i,ly_col,x_col,start_mark)
        if not res_one[0]:
            break
        res_list.append(res_one)
        i = res_one[0]
        start_mark = not start_mark
        pass
    return res_list

主方法

def enter_main(daily_file_path,start_date,end_date):
    # 1 截取要计算的时间区间对应的日数据
    df = pd.read_csv(daily_file_path,encoding='utf-8')
    df['o_date'] = pd.to_datetime(df['tradeDate'])
    df = df.loc[(df['o_date']>=start_date) & (df['o_date']<=end_date)].copy()
    df = df.loc[df['openPrice']>0].copy()
    df['o'] = df['openPrice'] * df['accumAdjFactor']
    df['c'] = df['closePrice'] * df['accumAdjFactor']
    df['h'] = df['highestPrice'] * df['accumAdjFactor']
    df['l'] = df['lowestPrice'] * df['accumAdjFactor']

    df = df.loc[:,['tradeDate','o','c','h','l']].copy()

    # 2 逆序，并设置索引字段
    df['i_row'] = [i for i in range(len(df))]
    i_row_list = df['i_row'].values.tolist()
    i_row_list.reverse()
    df['i_row_r'] = i_row_list
    h_list = df['h'].values.tolist()
    h_list.reverse()
    df['hr'] = h_list
    l_list = df['l'].values.tolist()
    l_list.reverse()
    df['lr'] = l_list

    # 3 从最新日期往前寻找所有转折点，即所有的峰谷值
    res_list = []
    i_len = len(i_row_list)
    p_first = caculate_turning_point(df,i_len,'hr','i_row_r',True)
    l_first = caculate_turning_point(df,i_len,'lr','i_row_r',False)
    if p_first[0]<l_first[0]:
        # 第一个
        res_list.append(l_first)
        res_list.append(p_first)
        # 谷开始
        res_list00 = circle_find(False, p_first[0], df, 'hr', 'lr', 'i_row_r')
        pass
    else:
        # 第一个
        res_list.append(p_first)
        res_list.append(l_first)
        # 峰开始
        res_list00 = circle_find(True, l_first[0], df, 'hr', 'lr', 'i_row_r')
        pass

    res_list.extend(res_list00)
    df_pv = pd.DataFrame(columns=['x','y','mark'],data=res_list)

    return df_pv.loc[:,['x','y']].values.tolist()

调用计算：

if __name__ == '__main__':
    file_path = r'D:/600959.csv'
    res = enter_main(file_path,'2023-01-01','2023-07-31')
    print(res)

[[132.0, 3.43], [132.0, 3.36], [131.0, 3.43], [130.0, 3.34], [127.0, 3.41], [126.0, 3.28], [124.0, 3.39], [122.0, 3.24], [118.0, 3.4], [115.0, 3.26], [107.0, 3.85], [106.0, 3.68], [105.0, 3.76], [102.0, 3.59], [100.0, 3.75], [96.0, 3.35], [95.0, 3.46], [94.0, 3.34], [93.0, 3.45], [92.0, 3.36], [88.0, 3.72], [87.0, 3.6], [81.0, 4.37], [76.0, 3.45], [73.0, 3.73], [72.0, 3.39], [71.0, 3.53], [69.0, 3.38], [66.0, 3.59], [64.0, 3.32], [63.0, 3.4], [62.0, 3.31], [60.0, 3.51], [58.0, 3.28], [55.0, 3.41], [53.0, 3.31], [51.0, 3.37], [50.0, 3.25], [49.0, 3.31], [47.0, 3.19], [45.0, 3.27], [44.0, 3.15], [42.0, 3.25], [41.0, 3.16], [39.0, 3.27], [34.0, 3.06], [32.0, 3.13], [31.0, 3.1], [30.0, 3.16], [28.0, 3.07], [27.0, 3.2], [26.0, 3.08], [25.0, 3.14], [24.0, 3.09], [23.0, 3.15], [22.0, 3.06], [21.0, 3.12], [19.0, 3.04], [18.0, 3.08], [15.0, 2.99], [14.0, 3.04], [12.0, 2.94], [11.0, 2.98], [10.0, 2.93], [9.0, 3.0], [7.0, 2.91], [5.0, 3.01], [4.0, 2.95], [2.0, 3.05]]