#!/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)
Done is better than perfect.