风控领域特征工程

# 创建示例数据集
data = pd.DataFrame({
   "CustomerID": [1, 2, 3, 4, 5, 6, 7, 8, 9, 10],
   "LastPurchaseDate": ['2023-01-01', '2023-02-15', '2023-03-20', 
                        '2023-04-10', '2023-05-05', '2023-06-12', 
                        '2023-07-08', '2023-08-22', '2023-09-30', '2023-10-15'],
   'Frequency': [2, 5, 1, 3, 2, 4, 1, 6, 3, 5],
   'Monetary': [100, 250, 50, 150, 120, 200, 40, 300, 180, 220]
})

# 转换日期列为日期类型
data['LastPurchaseDate'] = pd.to_datetime(data['LastPurchaseDate'])

# 计算Recency，即最近一次购买距今的天数
data['Recency'] = (dt.datetime.now() - data['LastPurchaseDate']).dt.days

# 展示RFM特征结果
print(data[['CustomerID', 'Recency', 'Frequency', 'Monetary']])

import pandas as pd
from sklearn.preprocessing import OneHotEncoder

# 示例数据
train_all_df = pd.DataFrame({
    "cust_gender": ["男", "男", "女", "女", "男"]
})

# 转换前的数据查看
print("转换前:", train_all_df["cust_gender"].head())

# 应用One-Hot编码
encoder = OneHotEncoder(sparse=False)
encoded_features = encoder.fit_transform(train_all_df[['cust_gender']])

# 将编码后的数据转换为DataFrame，并指定列的数据类型为整数
train_all_df2 = pd.DataFrame(encoded_features, columns=encoder.get_feature_names_out())
train_all_df2 = train_all_df2.astype(int)

# 转换后的数据查看
print("转换后:\n", train_all_df2.head())

import pandas as pd
from sklearn.preprocessing import LabelEncoder

# 示例数据
train_all_df = pd.DataFrame({
    "cust_gender": ["男", "男", "女", "女", "男"]
})

# 转换前的数据查看
print("转换前:", train_all_df["cust_gender"].head())

# 初始化标签编码器并转换数据
label_encoder = LabelEncoder()
train_all_df['cust_gender_encoded'] = label_encoder.fit_transform(train_all_df['cust_gender'])

# 转换后的数据查看
print("转换后:", train_all_df[['cust_gender', 'cust_gender_encoded']].head())

from sklearn.cluster import KMeans
import pandas as pd

# 创建示例数据集
data = pd.DataFrame({
    'ID': ['A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I'],
    "Feature1": [1, 2, 3, 8, 9, 10, 15, 16, 17],
    "Feature2": [5, 6, 7, 12, 13, 14, 1, 2, 3]
})

# 选择用于聚类的特征列
features = ['Feature1', 'Feature2']

# 使用K均值聚类算法
# 这里假设知道数据应该被分为3个聚类
kmeans = KMeans(n_clusters=3, random_state=42)

# 仅对特征列进行聚类，不包括 'ID' 列
data['Cluster'] = kmeans.fit_predict(data[features])

# 打印结果
print(data)

from sklearn.ensemble import IsolationForest
import pandas as pd

# 创建示例数据集
data = pd.DataFrame({
    "Feature1": [1, 2, 3, 8, 9, 10, 15, 16, 17, 100],
    "Feature2": [5, 6, 7, 12, 13, 14, 1, 2, 3, 200]
})

# 使用 Isolation Forest 算法进行异常值检测
isolation_forest = IsolationForest(contamination=0.01, random_state=42)

# 训练模型
isolation_forest.fit(data)

# 预测数据点是否为异常值，-1 表示异常值，1 表示正常值
outlier_scores = isolation_forest.predict(data)

# 将预测结果添加到数据框中
data['Outlier_Score'] = outlier_scores

# 打印结果
print(data)