通过企业维度权重归一化及领域匹配实现企业排名

一、pandas实现数据处理

1.pandas读取数据

import pandas as pd
io = r'D:\study\多标签企业数据（推荐用）.xlsx'
data = pd.read_excel(io, sheet_name = 0)
data.head(10)

结果如下图： 

 发现字段里面出现冗余数据‘-’，故尝试用0替换掉（上图数据量有限，未展示完全）

2.数据清洗

import re
registered_capital = data['注册资本']
capital = []
for i in registered_capital:
    m = str(i).replace('-','0')               #匹配到“-”并替换为“0”
    now = re.match("\d*", m).group()          #group(0) 同 group() 就是匹配正则表达式整体结果
    capital.append(now)
data['注册资本新'] = pd.DataFrame(capital)     #数据存入Excel列
data.to_excel(io)                             #数据保存

3.货币按汇率转换，并写入Excel

import regex
new_money = []
for item in data['注册资本新']:
    m1=regex.compile('\d+\.?\d*')
    money=float(m1.findall(item)[0])
    print(money)
    if '人民币' in item: money = money  
    if '美' in item: money = money * 6.4
    if '欧' in item:money = money * 7.8
    if '英' in item:money = money * 9.1
    if '法' in item:money = money * 7.1
    if '德' in item:money = money * 10.97
    if '加元' in item:money = money * 5.3
    if '新加坡' in item:money = money * 4.82
    if '港' in item:money = money * 0.82
    if '台币' in item:money = money * 0.232
    money2 = str(round(money,1))              #取小数点后一位
    new_money.append(money2)
data['汇率转换后的注册资本'] = pd.DataFrame(new_money)
data.to_excel(io)

二.企业经营领域划分

根据企业所属行业和经营范围匹配领域知识库，划分领域

import os
import numpy as np
import pandas as pd
class RecommendTo:
    def __init__(self):
        self.projectfilename = "D:/智慧园区20210426数据整合/多标签企业数据（推荐用）.xlsx"
        self.__keyworddict = self.test02()
        projs = self.test()
        projs = self.__evaluate_projects(projs)
        self.myprojs = projs
    def test02(self):
        dir = "D:/NoteBook/area_keywords/"
        keywordfilenames = []
        for filename in os.listdir(dir):
            if filename.endswith(".csv"):
                keywordfilenames.append(filename)
            else:
                continue
        keyworddict = {}
        for fname in keywordfilenames:
            keyworddict[fname.split('.')[0]] = np.concatenate(pd.read_csv(dir + fname).values)
        return keyworddict
    def test(self):
        projectfilename = self.projectfilename
        dir = "D:/NoteBook/area_keywords/"
        projs = pd.read_excel(projectfilename)
        return projs
    def __evaluate_projects(self, projs):
        # evaluate projects
        npprojs = projs.values
        projs_strs = []
        for item in npprojs:
            s = ""
            for area, keywords in self.__keyworddict.items():
                count = 0
                for keyword in keywords:
                    try:
                        # print(item[2],item[3])
                        if item[9].find(keyword) != -1:
                            count += 1
                        if item[12].find(keyword) != -1:
                            count += 1
                    except:
                        pass
                if count != 0:
                    s += area + ":" + str(count) + ","
            s = s.strip(",")
            projs_strs.append(s)
        projs["领域符合度"] = pd.DataFrame(projs_strs)
        projs.to_excel(self.projectfilename)
        return projs
RecommendTo()

三.数据展示

data.head()

 1.领域符合度计算

import re
usedate = data2['领域符合度']
shuju = []
for i in usedate:
    try :
        result = i.split(':')
        a=len(result)-1
        shuju.append(a)
    except:
        a = 0
        shuju.append(a)
data['领域数量'] = pd.DataFrame(shuju)
data.to_excel(io)

2.数据描述

data2.describe()

可以发现参保人数并没有展示出来，去excel发现数值需转换为数字

 数据清理，替换掉“-”

import re
insured = data2['参保人数']
insured_people = []
for i in insured:
    m = str(i).replace('-','0')                #匹配到“-”并替换为“0”
    now = re.match("\d*", m).group()           #group(0) 同 group() 就是匹配正则表达式整体结果
    insured_people.append(now)
data2['参保人数新'] = pd.DataFrame(insured_people)    #数据存入Excel列
data2.to_excel(r'D:\\多标签企业数据（推荐用）.xlsx')     

round(data2.describe(),1) #数据描述，并保留一位小数

由于极大值和极小值相差较大，且平均数受影响较大，于是考虑采用中位数，分别对注册资本和参保人数进行数据缺失补全。

注册资金根据中位数缺失数据补全：

import re
shiyong=data2['注册资本新']
shuju = []
for i in shiyong:
    if i == 0:
        i = 1000
        shuju.append(i)
    else:
        i = i
        shuju.append(i)
# print(shuju)
data2['注册资本补全'] = pd.DataFrame(shuju)
data2.to_excel(r'D:\多标签企业数据（推荐用）.xlsx')

参保人数根据中位数缺失数据补全：

import re
shiyong=data2['参保人数新']
shuju = []
for i in shiyong:
    if i == 0:
        i = 3
        shuju.append(i)
    else:
        i = i
        shuju.append(i)
# print(shuju)
data2['参保人数补全'] = pd.DataFrame(shuju)
data2.to_excel(r'D:/多标签企业数据（推荐用）.xlsx')

当前数据格式

data2.head(10)

四.数据权重归一化

from sklearn.preprocessing import MinMaxScaler
x1 = data[['注册资本补全']]
x2 = data[['参保人数补全']]
x3 = data[['领域数量']]
min_max_scaler = MinMaxScaler()
X_train_minmax1 = min_max_scaler.fit_transform(x1)#注册资本归一化后的结果
X_train_minmax2 = min_max_scaler.fit_transform(x2)#参保人数归一化后的结果
X_train_minmax3 = min_max_scaler.fit_transform(x3)#领域数量归一化后的结果
score = 10*(X_train_minmax1+X_train_minmax2+X_train_minmax3)+84  #数据加权化到[0,100]
round(score.max(),1)           #保留一位小数

输出：99.0

 score.min() 的输出为84.0，就是说整体权重划分到[84.100]，在图上展示结果

import seaborn as sns
ax = sns.distplot(score)
ax.set_title('score')

 归一化和加权得分写入Excel

data['score'] = pd.DataFrame(score)
data.to_excel(io)

data.head()

得到的最终结果，如下图所示

以上就是对通过数据清洗、补全、归一化、加权得到的最终企业综合评分。