数据分析-task 2（特征工程）

task2 任务要求

特征衍生

特征挑选：分别用IV值和随机森林等进行特征选择

……以及你能想到特征工程处理

使用IV值特征选择

def calcWOE(dataset, col, target):

    # 对特征进行统计分组

    subdata = df(dataset.groupby(col)[col].count())

    # 每个分组中响应客户的数量

    suby = df(dataset.groupby(col)[target].sum())

    # subdata 与 suby 的拼接

    data = df(pd.merge(subdata, suby, how='left', left_index=True, right_index=True))

    # 相关统计，总共的样本数量total，响应客户总数b_total，未响应客户数量g_total

    b_total = data[target].sum()

    total = data[col].sum()

    g_total = total - b_total

    # WOE公式

    data["bad"] = data.apply(lambda x:round(x[target]/b_total, 100), axis=1)

    data["good"] = data.apply(lambda x:round((x[col] - x[target])/g_total, 100), axis=1)

    data["WOE"] = data.apply(lambda x:log(x.bad / x.good), axis=1)

    return data.loc[:, ["bad", "good", "WOE"]]

def calcIV(dataset):

    dataset["IV"] = dataset.apply(lambda x:(x["bad"] - x["good"]) * x["WOE"], axis=1)

    IV = sum(dataset["IV"])

    return IV

y = data.status

x= data.drop('status', axis=1)

col_list = [col for col in data.drop(labels=['Unnamed: 0','status'], axis=1)]

data_IV = df()

fea_iv = []

for col in col_list:

    col_WOE = calcWOE(data, col, "status")

    # 删除nan、inf、-inf

#    col_WOE = col_WOE[~col_WOE.isin([np.nan, np.inf, -np.inf]).any(1)]

    col_IV = calcIV(col_WOE)

    if col_IV > 0.1:

        data_IV[col] = [col_IV]

        fea_iv.append(col)

data_IV.to_csv('feature.csv', index=0,encoding='gbk')

print(fea_iv)

2. 使用随机森林特征选择

from sklearn.ensemble import RandomForestClassifier

rfc = RandomForestClassifier(random_state=2018)

rfc.fit(x, y)

importance = pd.Series(rfc.feature_importances_, index=x.columns).sort_values(ascending=False)

rfc_result = importance[: 20].index.tolist()

print(rfc_result)