多label实现准确率和召回率

#!/usr/bin/env python
# -*- coding: utf-8 -*-
"""
    @File : recall.py
    @Author : 郭凯锋
    @Time : 2020/1/12 17:57
    @Software : PyCharm
    @Git-Hub : daguonice
    @博客园: https://www.cnblogs.com/daguonice/
    
"""
import pandas as pd
import numpy as np
from sklearn.metrics import recall_score


def right_or_wrong(ypred_ele, ytrue_ele):
    if ypred_ele in ytrue_ele:
        return True
    else:
        return False


def get_single_score(y_true, y_pred):
    TP = np.sum(np.logical_and(np.equal(y_true, 1), np.equal(y_pred, 1)))
    FP = np.sum(np.logical_and(np.equal(y_true, 0), np.equal(y_pred, 1)))
    TN = np.sum(np.logical_and(np.equal(y_true, 1), np.equal(y_pred, 0)))
    FN = np.sum(np.logical_and(np.equal(y_true, 0), np.equal(y_pred, 0)))
    recall = [TP, (TP + FN)]
    accuracy = [(TP + TN), len(y_pred)]
    return accuracy, recall


def get_multi_score(y_true, y_pred):
    list_score = []
    for ele in ['相互宝', '健康险', '保险线', '花呗']:
        y_true_temp = np.zeros(len(y_true))
        y_pred_temp = np.zeros(len(y_pred))
        for idx in range(len(y_true)):
            if y_pred[idx] == ele:
                y_pred_temp[idx] = 1.
            if right_or_wrong(ele, y_true[idx]):
                y_true_temp[idx] = 1.
        accuracy, recall = get_single_score(y_true_temp, y_pred_temp)  # 变成百分制
        list_score.append([accuracy, recall])
    return list_score  # 返回一个三维列表， 分别是相互宝、健康险、保险线的准确率和召回率的分数


def fraction2decimal(result):
    res = []
    for ele in result:
        list_ele = []
        for lab in ele:
            list_lab = []
            for score in lab:
                if score[1] == 0:
                    list_lab.append(0.0)
                else:
                    list_lab.append(round(score[0] / score[1], 2))
            list_ele.append(list_lab)
        res.append(list_ele)
    return res


def func(y_true, y_pred):
    # y_true是一个一维列表， y_pred是一个二维列表
    if len(y_true) != len(y_pred):
        raise Exception("The two input lengths are inconsistent")
    length = len(y_true)
    list_score = get_multi_score(y_true, y_pred)
    result = []
    temp_list = list_score
    for idx in range(length):
        list_score = temp_list
        if len(y_true[idx]) == 1 and y_pred[idx] not in y_true[idx]:
            list_score[1][1][1] += 1
            list_score[0][0][1] += 1
            result.append(list_score)
        elif len(y_true[idx]) == 2 and y_pred[idx] in y_true[idx]:
            list_score[0][0][0] += 1
            list_score[0][0][1] += 1
            list_score[0][1][0] += 1
            list_score[0][1][1] += 1
            result.append(list_score)
        elif len(y_true[idx]) == 2 and y_pred[idx] not in y_true[idx]:
            list_score[1][1][1] += 1
            list_score[2][1][1] += 1
            list_score[0][0][1] += 1
            result.append(list_score)
        else:
            result.append(list_score)
    result = fraction2decimal(result)
    return result


if __name__ == '__main__':
    alist = ['相互宝', '相互宝', '相互宝', '健康险', '保险线', '花呗']
    blist = ['健康险', ['健康险', '相互宝'], ['健康险', '保险线'], ['相互宝'], ['保险线'], ['花呗']]
    res = func(blist, alist)
    # res = pd.DataFrame(res, columns=['相互宝', '健康险', '保险线', '花呗'])
    # pd.set_option('display.max_columns', None)
    # pd.set_option('display.max_rows', None)
    # pd.set_option('display.width', 100000)
    # pd.set_option('display.unicode.east_asian_width', True)
    print()
    print(res)