用戶流失預警項目實戰(zhàn)[大數(shù)據(jù)培訓]

#TODO:用戶流失預警
#1.導包
from __future__ import division
import pandas as pd
import numpy as np
#2.加載數(shù)據(jù)與觀察數(shù)據(jù)
churn_df = pd.read_csv('UserDrain_data/churn.csv')
col_names = churn_df.columns.tolist()
# print("Column names:\n",col_names)
# print(churn_df.shape) #(3333, 21)
# Column names:['State', 'Account Length', 'Area Code', 'Phone', "Int'l Plan", 'VMail Plan',
'VMail Message', 'Day Mins', 'Day Calls', 'Day Charge', 'Eve Mins', 'Eve Calls', 'Eve Charge',
'Night Mins', 'Night Calls', 'Night Charge', 'Intl Mins', 'Intl Calls', 'Intl Charge', 'CustServ
Calls', 'Churn?']
to_show = col_names[:6] + col_names[-6:] #前6 列與后6 列
# print(len(to_show))#共12 列
# print ("\nSample data:\n",churn_df[to_show].head(6))
# State Account Length Area Code ... Intl Charge CustServ Calls Churn?
# 0 KS 128 415 ... 2.70 1 False.
# 1 OH 107 415 ... 3.70 1 False.
#2.1 類別編碼
churn_result = churn_df['Churn?']
y = np.where(churn_result == 'True.',1,0)
#2.2 刪除不需要對應(yīng)列數(shù)據(jù)
to_drop = ['State','Area Code','Phone','Churn?']
churn_feat_space = churn_df.drop(to_drop,axis=1)
#yes 或者no 需要轉(zhuǎn)化為布爾類型數(shù)據(jù)
yes_no_cols = ["Int'l Plan","VMail Plan"]
churn_feat_space[yes_no_cols] = churn_feat_space[yes_no_cols] == 'yes'
# print(churn_feat_space[yes_no_cols])
# Int'l Plan VMail Plan
# 0 False True
# 1 False True
# 獲取數(shù)據(jù)屬性名
features = churn_feat_space.columns
# print("churn_feat_space:\n",churn_feat_space.head())
# Account Length Int'l Plan ... Intl Charge CustServ Calls
# 0 128 False ... 2.7 1
# 1 107 False ... 3.7 1
#將dataframe 轉(zhuǎn)化為ndarray 數(shù)組，同時數(shù)組中的元素類型為float 類型
#對應(yīng)的布爾類型的值，True 為1，False 為0
X = churn_feat_space.as_matrix().astype(np.float)
np.set_printoptions(threshold=np.NaN)
# print("churn_feat_space.as_matrix().astype(np.float):\n",X)
# [[128. 0. 1.... 3. 2.7 1.]
# [107. 0. 1.... 3. 3.7 1.]
# 2.3 數(shù)據(jù)標準化
from sklearn.preprocessing import StandardScaler
scaler = StandardScaler()
X = scaler.fit_transform(X)
# print("Feature space holds %d observations and %d features" % X.shape) #3333 行17 列
# Feature space holds 3333 observations and 17 features
# print("Unique target labels:", np.unique(y))# [0 1]
# print(X[0])
#標準化之后的第一行數(shù)據(jù)
#[ 0.67648946 -0.32758048 1.6170861 1.23488274 1.56676695 0.47664315
# 1.56703625 -0.07060962 -0.05594035 -0.07042665 0.86674322 -0.46549436
# 0.86602851 -0.08500823 -0.60119509 -0.0856905 -0.42793202]
# print(len(y[y == 0])) #2850
#3.KFold K 折交叉驗證
from sklearn.cross_validation import KFold
def run_cv(X,y,clf_class,**kwargs):
#創(chuàng)建kfold 對象
kf=KFold(len(y),n_folds=5,shuffle=True)
y_pred=y.copy()
#迭代
count=0
for train_index,test_index in kf:
count=count+1
#y 對應(yīng)的標簽數(shù)量：3333
# print("train_index 數(shù)量:",len(train_index)) #train_index 數(shù)量: 2666
# print("test_index 數(shù)量:", len(test_index)) # test_index 數(shù)量: 667
# print(test_index)
X_train,X_test=X[train_index],X[test_index]
y_train=y[train_index]
#初始化一個分類器模型
clf=clf_class(**kwargs)
clf.fit(X_train,y_train)
y_pred[test_index] = clf.predict(X_test)
# print("迭代次樹:", count)#5
return y_pred
#4.建模
from sklearn.svm import SVC
from sklearn.ensemble import RandomForestClassifier as RF
from sklearn.neighbors import KNeighborsClassifier as KNN
def accuracy(y_true,y_pred):
# NumPy interprets True and False as 1. and 0.
return np.mean(y_true == y_pred)
#4.1 SVM / RF /KNN 三種算法預測準確率
print("Support vector machines:","%.3f" % accuracy(y, run_cv(X,y,SVC)))
print("Random forest:","%.3f" % accuracy(y, run_cv(X,y,RF)))
print("K-nearest-neighbors:","%.3f" % accuracy(y,run_cv(X,y,KNN)))