Topsis打分评级

import pandas as pd
import numpy as np


def topsis(data, weight=None):
    data = data / np.sqrt((data**2).sum())

    ideal_solution = data.max()
    negative_ideal_solution = data.min()

    distance_to_ideal_solution = np.sqrt(
        ((data - ideal_solution)**2 * weight).sum(axis=1))
    distance_to_negative_ideal_solution = np.sqrt(
        ((data - negative_ideal_solution)**2 * weight).sum(axis=1))

    similarity = distance_to_negative_ideal_solution / (
        distance_to_ideal_solution + distance_to_negative_ideal_solution)

    rank = similarity.rank(method='min', ascending=False)

    return rank, similarity


data = pd.read_excel('data1-3（睡醒次数+1）.xlsx')
sleep_quality_data = data[["整晚睡眠时间", "睡醒次数", "入睡方式"]]

# 权重
weight = np.array([1, 1, 1])

rank, similarity = topsis(sleep_quality_data, weight)

data["相似度"] = similarity
data["等级"] = rank

data["睡眠质量"] = pd.qcut(data["相似度"], 4, labels=["差", "中", "良", "优"])

data.to_excel('data2（Topsis评分评级）.xlsx')

唯一的缺点是我需要一个服从正态分布的等级分布，但是结果是四个等级人数差不多。所以后来我直接自己定义评分标准了。