【机器学习项目】电信行业的客户流失预测模型

案例：电信行业的客户流失预测模型

项目背景：

在电信行业，顾客可以从各种服务提供商中选择。顾客流失被定义为顾客停止与公司或服务进行业务往来的情况。

项目任务是使用提供的其余数据建立一个预测客户流失的模型。

数据集介绍：
此数据集包括电信公司的客户数据，包括服务使用情况、人口统计数据以及客户是否流失。

有需要下载数据练习的， 可点击这篇文章 发送文末暗号到公众号领取。

具体代码

import pandas as pd
import numpy as np
from sklearn.model_selection import train_test_split, cross_val_score, KFold
from sklearn.preprocessing import StandardScaler, OneHotEncoder,LabelEncoder
from sklearn.impute import SimpleImputer
from sklearn.pipeline import Pipeline
from sklearn.ensemble import RandomForestClassifier
from xgboost import XGBClassifier
from catboost import CatBoostClassifier
from sklearn.metrics import accuracy_score
from sklearn.compose import ColumnTransformer
import matplotlib.pyplot as plt
import seaborn as sns


data==pd.read_csv('WA_Fn-UseC_-Telco-Customer-Churn.csv')
target='Churn'

y = data[target]
X=data.drop(target,axis=1)

if not all(isinstance(val, int) for val in y):
    label_encoder = LabelEncoder()
    y = label_encoder.fit_transform(y)
else:
    y

numerical_cols = X.select_dtypes(include=[np.number]).columns.tolist()
categorical_cols = X.select_dtypes(include=[object]).columns.tolist()


# Preprocessor Pipeline setup
numerical_transformer = Pipeline([
    ('imputer', SimpleImputer(strategy='mean')),
    ('scaler', StandardScaler())
])

categorical_transformer = Pipeline([
    ('imputer', SimpleImputer(strategy='most_frequent')),
    ('onehot', OneHotEncoder(handle_unknown='ignore'))
])

preprocessor = ColumnTransformer(
    transformers=[
        ('num', numerical_transformer, numerical_cols),
        ('cat', categorical_transformer, categorical_cols)
    ])


# train test split
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

# Model training pipeline
models = {
    'RandomForest': RandomForestClassifier(),
    'XGBoost': XGBClassifier(),
    'CatBoost': CatBoostClassifier(verbose=0)
}

pipeline_results = {}

for name, model in models.items():
    pipeline = Pipeline([
        ('preprocessor', preprocessor),
        ('classifier', model)
    ])
    
    kf = KFold(n_splits=5, shuffle=True, random_state=42)
    cv_results = cross_val_score(pipeline, X_train, y_train, cv=kf, scoring='accuracy')
    pipeline_results[name] = cv_results.mean()

# print results 
print(pipeline_results)    

三种模型的准确度都在80%左右，其中catboost 表现最好，RF其次，XGB最后。

# Select the best model
best_model_name = max(pipeline_results, key=pipeline_results.get)
best_pipeline = Pipeline([
    ('preprocessor', preprocessor),
    ('classifier', models[best_model_name])
])
best_pipeline.fit(X_train, y_train)

# Printing the accuracy of the best model
y_pred = best_pipeline.predict(X_test)
accuracy = accuracy_score(y_test, y_pred)
print(f'Best Model: {best_model_name}')
print(f'Accuracy: {accuracy}')

在测试集上的结果显示准确率是81%左右，使用的是catboost模型

# Get column names
numerical_cols = preprocessor.named_transformers_['num'].get_feature_names_out(input_features=numerical_cols)
categorical_cols_encoded = preprocessor.named_transformers_['cat'].named_steps['onehot'].get_feature_names_out(input_features=categorical_cols)
all_columns = list(numerical_cols) + list(categorical_cols_encoded)

# get feature importance
feature_importance = model.feature_importances_

# dataframe
feature_importance_df = pd.DataFrame({'Feature': all_columns, 'Importance': feature_importance})
feature_importance_df = feature_importance_df.sort_values(by='Importance', ascending=False)
feature_importance_df= feature_importance_df.head(10)


# visualise top 10 features
plt.figure(figsize=(10, 6))
sns.barplot(x='Importance', y='Feature', data=feature_importance_df)
plt.title('Feature Importance')
plt.xlabel('Importance')
plt.ylabel('Feature')
plt.show()

总结

可以进一步通过特征选择、参数优化等方式，提高模型的性能比如：

• 确定相关特征： 确定与流失分析相关的主要数据特征
• 相关性分析： 分析特征之间的相关性，以确定它们对流失的影响程度
• 特征重要性： 分析特征的重要性，了解哪些因素对于预测流失最为关键

每个企业的流失分析过程可能会有所不同，具体的方法和模型选择取决于业务的性质、可用的数据以及分析的目标。