# -*-coding:utf-8-*-
"""
Author: Thinkgamer
Desc:
代码8-1： 第八章 GBDT模型 电信客户流失预测
"""
from sklearn.model_selection import train_test_split
from sklearn.ensemble import GradientBoostingClassifier
from sklearn import metrics
from sklearn.metrics import mean_squared_error
import pandas as pd
import os
class ChurnPredWithGBDT:
def __init__(self):
self.file = "data/telecom-churn/telecom-churn-prediction-data.csv"
self.data = self.feature_transform()
self.train, self.test = self.split_data()
# 空缺值以0值填充
def isNone(self, value):
if value == " " or value is None:
return "0.0"
else:
return value
# 特征转换
def feature_transform(self):
if not os.path.exists("./data/new-churn.csv"):
print("Start Feature Transform ...")
# 定义特征转换字典
feature_dict = {
"gender": {"Male": "1", "Female": "0"},
"Partner": {"Yes": "1", "No": "0"},
"Dependents": {"Yes": "1", "No": "0"},
"PhoneService": {"Yes": "1", "No": "0"},
"MultipleLines": {"Yes": "1", "No": "0", "No phone service": "2"},
"InternetService": {"DSL": "1", "Fiber optic": "2", "No": "0"},
"OnlineSecurity": {"Yes": "1", "No": "0", "No internet service": "2"},
"OnlineBackup": {"Yes": "1", "No": "0", "No internet service": "2"},
"DeviceProtection": {"Yes": "1", "No": "0", "No internet service": "2"},
"TechSupport": {"Yes": "1", "No": "0", "No internet service": "2"},
"StreamingTV": {"Yes": "1", "No": "0", "No internet service": "2"},
"StreamingMovies": {"Yes": "1", "No": "0", "No internet service": "2"},
"Contract": {"Month-to-month": "0", "One year": "1", "Two year": "2"},
"PaperlessBilling": {"Yes": "1", "No": "0"},
"PaymentMethod": {
"Electronic check": "0",
"Mailed check": "1",
"Bank transfer (automatic)": "2",
"Credit card (automatic)": "3",
},
"Churn": {"Yes": "1", "No": "0"},
}
fw = open("data/new_churn.csv", "w")
fw.write(
"customerID,gender,SeniorCitizen,Partner,Dependents,tenure,PhoneService,MultipleLines,"
"InternetService,OnlineSecurity,OnlineBackup,DeviceProtection,TechSupport,StreamingTV,"
"StreamingMovies,Contract,PaperlessBilling,PaymentMethod,MonthlyCharges,TotalCharges,Churn\n"
)
for line in open(self.file, "r").readlines():
if line.startswith("customerID"):
continue
customerID, gender, SeniorCitizen, Partner, Dependents, tenure, PhoneService, MultipleLines, \
InternetService, OnlineSecurity, OnlineBackup, DeviceProtection, TechSupport, StreamingTV, \
StreamingMovies, Contract, PaperlessBilling, PaymentMethod, MonthlyCharges, TotalCharges, Churn \
= line.strip().split(",")
_list = list()
_list.append(customerID)
_list.append(self.isNone(feature_dict["gender"][gender]))
_list.append(self.isNone(SeniorCitizen))
_list.append(self.isNone(feature_dict["Partner"][Partner]))
_list.append(self.isNone(feature_dict["Dependents"][Dependents]))
_list.append(self.isNone(tenure))
_list.append(self.isNone(feature_dict["PhoneService"][PhoneService]))
_list.append(self.isNone(feature_dict["MultipleLines"][MultipleLines]))
_list.append(
self.isNone(feature_dict["InternetService"][InternetService])
)
_list.append(
self.isNone(feature_dict["OnlineSecurity"][OnlineSecurity])
)
_list.append(self.isNone(feature_dict["OnlineBackup"][OnlineBackup]))
_list.append(
self.isNone(feature_dict["DeviceProtection"][DeviceProtection])
)
_list.append(self.isNone(feature_dict["TechSupport"][TechSupport]))
_list.append(self.isNone(feature_dict["StreamingTV"][StreamingTV]))
_list.append(
self.isNone(feature_dict["StreamingMovies"][StreamingMovies])
)
_list.append(self.isNone(feature_dict["Contract"][Contract]))
_list.append(
self.isNone(feature_dict["PaperlessBilling"][PaperlessBilling])
)
_list.append(self.isNone(feature_dict["PaymentMethod"][PaymentMethod]))
_list.append(self.isNone(MonthlyCharges))
_list.append(self.isNone(TotalCharges))
_list.append(feature_dict["Churn"][Churn])
fw.write(",".join(_list))
fw.write("\n")
return pd.read_csv("data/new_churn.csv")
else:
return pd.read_csv("data/new_churn.csv")
# 数据集拆分为训练集和测试集
def split_data(self):
train, test = train_test_split(
self.data,
test_size=0.1,
random_state=40
)
return train, test
# 调用skleran进行模型训练
def train_model(self):
print("Start Train Model ... ")
lable = "Churn"
ID = "customerID"
x_columns = [x for x in self.train.columns if x not in [lable, ID]]
x_train = self.train[x_columns]
y_train = self.train[lable]
gbdt = GradientBoostingClassifier(
learning_rate=0.1, n_estimators=200, max_depth=5
)
gbdt.fit(x_train, y_train)
return gbdt
# 模型评估
def evaluate(self, gbdt):
lable = "Churn"
ID = "customerID"
x_columns = [x for x in self.test.columns if x not in [lable, ID]]
x_test = self.test[x_columns]
y_test = self.test[lable]
y_pred = gbdt.predict_proba(x_test)
new_y_pred = list()
for y in y_pred:
# y[0] 表示样本label=0的概率 y[1]表示样本label=1的概率
new_y_pred.append(1 if y[1] > 0.5 else 0)
mse = mean_squared_error(y_test, new_y_pred)
print("MSE: %.4f" % mse)
accuracy = metrics.accuracy_score(y_test.values, new_y_pred)
print("Accuracy : %.4g" % accuracy)
auc = metrics.roc_auc_score(y_test.values, new_y_pred)
print("AUC Score : %.4g" % auc)
if __name__ == "__main__":
pred = ChurnPredWithGBDT()
gbdt = pred.train_model()
print(gbdt)
pred.evaluate(gbdt)给出具体解决方法C:\Users\Administrator\PycharmProjects\PythonProject1\.venv\Scripts\python.exe C:\Users\Administrator\Desktop\推荐算法\案例7-1-基于GBDT的排序方法\案例7-1-基于GBDT的排序方法\GBDT模型的电信客户流失预测.py
Start Feature Transform ...
Traceback (most recent call last):
File "C:\Users\Administrator\Desktop\推荐算法\案例7-1-基于GBDT的排序方法\案例7-1-基于GBDT的排序方法\GBDT模型的电信客户流失预测.py", line 150, in <module>
pred = ChurnPredWithGBDT()
File "C:\Users\Administrator\Desktop\推荐算法\案例7-1-基于GBDT的排序方法\案例7-1-基于GBDT的排序方法\GBDT模型的电信客户流失预测.py", line 18, in __init__
self.data = self.feature_transform()
~~~~~~~~~~~~~~~~~~~~~~^^
File "C:\Users\Administrator\Desktop\推荐算法\案例7-1-基于GBDT的排序方法\案例7-1-基于GBDT的排序方法\GBDT模型的电信客户流失预测.py", line 62, in feature_transform
for line in open(self.file, "r").readlines():
~~~~^^^^^^^^^^^^^^^^
FileNotFoundError: [Errno 2] No such file or directory: 'data/telecom-churn/telecom-churn-prediction-data.csv'
扫一扫
7896
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
