import pandas as pd
import numpy as np
from collections import Counter
from sklearn import preprocessing
from matplotlib import pyplot as plt
%matplotlib inline
import seaborn as sns 
plt.rcParams['font.sans-serif'] = ['SimHei'] # 中文字體設(shè)置-黑體
plt.rcParams['axes.unicode_minus'] = False # 解決保存圖像是負(fù)號(hào)'-'顯示為方塊的問(wèn)題
sns.set(font='SimHei') # 解決Seaborn中文顯示問(wèn)題

data=pd.read_excel('dummy.xlsx')

data.head() #查看長(zhǎng)啥樣
data.shape #查看數(shù)據(jù)的行列大小
data.describe()

#列級(jí)別的判斷,但凡某一列有null值或空的，則為真
data.isnull().any()

#將列中為空或者null的個(gè)數(shù)統(tǒng)計(jì)出來(lái)，并將缺失值最多的排前
total = data.isnull().sum().sort_values(ascending=False)
print(total)


#輸出百分比：
percent =(data.isnull().sum()/data.isnull().count()).sort_values(ascending=False)
missing_data = pd.concat([total, percent], axis=1, keys=['Total', 'Percent'])
missing_data.head(20)

import missingno
missingno.matrix(data)
data=data.dropna(thresh=data.shape[0]*0.5,axis=1) #至少有一半以上是非空的列篩選出來(lái)

#如果某一行全部都是na才刪除：
data.dropna(axis=0,how='all') 

#默認(rèn)情況下是只保留沒(méi)有空值的行
data=data.dropna(axis=0) 

#統(tǒng)計(jì)重復(fù)記錄數(shù)
data.duplicated().sum()
data.drop_duplicates()

data.columns
#第一步，將整個(gè)data的連續(xù)型字段和離散型字段進(jìn)行歸類
id_col=['姓名']
cat_col=['學(xué)歷','學(xué)校'] #這里是離散型無(wú)序，如果有序，請(qǐng)參考map用法，一些博客上有寫
cont_col=['成績(jī)','能力'] #這里是數(shù)值型
print (data[cat_col]) #這里是離散型的數(shù)據(jù)部分
print (data[cont_col])#這里是連續(xù)性數(shù)據(jù)部分

#計(jì)算出現(xiàn)的頻次
for i in cat_col:
  print (pd.Series(data[i]).value_counts())
  plt.plot(data[i])

#對(duì)于離散型數(shù)據(jù)，對(duì)其獲取啞變量
dummies=pd.get_dummies(data[cat_col])
dummies

#對(duì)于連續(xù)型數(shù)據(jù)的大概統(tǒng)計(jì)：
data[cont_col].describe()

#對(duì)于連續(xù)型數(shù)據(jù)，看偏度，一般大于0.75的數(shù)值做一個(gè)log轉(zhuǎn)化，使之盡量符合正態(tài)分布，因?yàn)楹芏嗄Ｐ偷募僭O(shè)數(shù)據(jù)是服從正態(tài)分布的
skewed_feats = data[cont_col].apply(lambda x: (x.dropna()).skew() )#compute skewness
skewed_feats = skewed_feats[skewed_feats > 0.75]
skewed_feats = skewed_feats.index
data[skewed_feats] = np.log1p(data[skewed_feats])
skewed_feats

#對(duì)于連續(xù)型數(shù)據(jù)，對(duì)其進(jìn)行標(biāo)準(zhǔn)化
scaled=preprocessing.scale(data[cont_col])
scaled=pd.DataFrame(scaled,columns=cont_col)
scaled

m=dummies.join(scaled)
data_cleaned=data[id_col].join(m)
data_cleaned

data_cleaned.corr()

#以下是相關(guān)性的熱力圖，方便肉眼看
def corr_heat(df):
  dfData = abs(df.corr())
  plt.subplots(figsize=(9, 9)) # 設(shè)置畫面大小
  sns.heatmap(dfData, annot=True, vmax=1, square=True, cmap="Blues")
  # plt.savefig('./BluesStateRelation.png')
  plt.show()
corr_heat(data_cleaned)

#取出與某個(gè)變量（這里指能力）相關(guān)性最大的前四個(gè)，做出熱點(diǎn)圖表示
k = 4 #number of variables for heatmap
cols = corrmat.nlargest(k, '能力')['能力'].index
cm = np.corrcoef(data_cleaned[cols].values.T)
sns.set(font_scale=1.25)
hm = sns.heatmap(cm, cbar=True, annot=True, square=True, fmt='.2f',   annot_kws={'size': 10}, yticklabels=cols.values,  xticklabels=cols.values)
plt.show()