机器学习项目实战基于Python实现的保险反欺诈预测源代码+数据集

import pandas as pd import numpy as np import matplotlib.pyplot as plt import datetime train = pd.read_csv('train.csv') test = pd.read_csv('test.csv') submission = pd.read_csv("submission.csv") data = pd.concat([train, test], axis=0, ignore_index=True) # 处理时序数据 data['policy_bind_date'] = pd.to_datetime(data['policy_bind_date'], format='%Y-%m-%d') data['incident_date'] = pd.to_datetime(data['incident_date'], format='%Y-%m-%d') data["year"] = data["policy_bind_date"].dt.year data["month"] = data["policy_bind_date"].dt.month data["year1"] = data["incident_date"].dt.year data["month1"] = data["incident_date"].dt.month startdate = datetime.datetime.strptime('2022-06-30', '%Y-%m-%d') data['time'] = data['incident_date'].apply(lambda x: startdate-x).dt.days data = data.drop(["incident_date"], axis=1) data = data.drop(["policy_bind_date"], axis=1) # 把所有的相同类别的特征编码为同一个值 numerical_features = [x for x in data.columns if data[x].dtype != object] category_features = [x for x in data.columns if data[x].dtype == object] def get_mapfunction(x): mapp = dict(zip(x.unique().tolist(), range(len(x.unique().tolist())))) def mapfunction(y): if y in mapp: return mapp[y] else: return -1 return mapfunction for i in category_features: data[i] = data[i].apply(get_mapfunction(data[i])) y_train_ = data["fraud"][:train.shape[0]] data = data.drop(["policy_id"], axis=1) data = data.drop(["fraud"], axis=1) x_train_ = data[:train.shape[0]] x_test_ = data[train.shape[0]:] x_train = np.array(x_train_) y_train = np.array(y_train_) x_test = np.array(x_test_) from lightgbm.sklearn import LGBMClassifier # 从sklearn库中导入网格调参函数 # from sklearn.model_selection import GridSearchCV # # 定义参数取值范围, 网格调参，它的基本思想是穷举搜索 # learning_rate = [0.1] # feature_fraction = [0.5] # num_leaves = [20, 24, 16] # max_depth = [-1, 3, 4, 5] # parameters = {'learning_rate': learning_rate, # 'feature_fraction':feature_fraction, # 'num_leaves': num_leaves, # 'max_depth': max_depth} # model = LGBMClassifier(n_estimators=50) # # 进行网格搜索 # clf = GridSearchCV(model, parameters, cv=5, scoring='accuracy', verbose=1, n_jobs=-1) # clf = clf.fit(x_train, y_train) # # 网格搜索后的最好参数为 # print(clf.best_params_) clf = LGBMClassifier(feature_fraction=0.5, learning_rate=0.1, max_depth=4, num_leaves=20) clf.fit(x_train, y_train) from sklearn.model_selection import KFold folds = KFold(n_splits=10, shuffle=True, random_state=2022) # 10折交叉验证 oof_xgb = np.zeros(len(train)) # 用于存放训练集的预测 predictions_xgb = np.zeros(len(test)) # 用于存放测试集的预测 for fold_, (trn_idx, val_idx) in enumerate(folds.split(x_train, y_train)): print("fold n°{}".format(fold_ + 1)) clf.fit(x_train[trn_idx], y_train[trn_idx]) oof_xgb[val_idx] = clf.predict(x_train[val_idx]) # 预测验证集 predictions_xgb += clf.predict(x_test) / folds.n_splits # 预测测试集，并且取平均 submission['fraud'] = predictions_xgb submission.to_csv('submit3.csv', index=False) from sklearn.metrics import accuracy_score print('acc=', accuracy_score(oof_xgb, y_train))