使用Python+Autogluon对“员工自评”进行机器学习建模分析

建模核心代码

#员工自评AutoML
from autogluon.tabular import TabularDataset, TabularPredictor
import warnings
warnings.filterwarnings('ignore')
train_data = TabularDataset(train_df2)

# 预测标签
label = '员工自评'

# 模型保存文件名
save_path = '../data/AUO-train/model/AM/worker5/worker5.pkl'

# 建立预测模型，verbosity(0~4)，默认为2就好

worker5 = TabularPredictor(label=label,path=save_path,verbosity=2)

# presets='best_quality'不考虑时间成本，追求最好模型
worker5.fit(train_data,presets='best_quality',num_bag_folds=5,num_bag_sets=1,num_stack_levels=1)

# 输出模型表现
worker5.leaderboard(silent=True)

训练结果

保留最优模型

# 删除其余模型（减少内存开销）
worker5.delete_models(models_to_keep='best')
# 输出最优模型
worker5.get_model_best()

保存模型

import sklearn.externals
import joblib

#保存模型
joblib.dump(worker5, '../data/AUO-train/model/AM/worker5/worker5_best_0705.pkl')

加载模型

#加载模型
import joblib
model_best=joblib.load('../data/AUO-train/model/AM/worker5/worker5_best.pkl')

使用模型进行预测

pre_test=model_best.predict(test_df2)
pre_test=pd.DataFrame(pre_test)
pre_test

模型评估

自定义模型评估函数

def good_or_bad_model(y_test, y_test_pred):
    from sklearn.metrics import explained_variance_score, \
    mean_absolute_error, mean_squared_error, \
    median_absolute_error, r2_score
    
    print(f'均方根误差为": {np.sqrt(mean_squared_error(y_test, y_test_pred))} ')
    
    print('均方误差为：', mean_squared_error(y_test, y_test_pred))
    
    print(f'平均绝对误差为: {mean_absolute_error(y_test, y_test_pred)}')
    
    print(f'R方值为: { r2_score(y_test, y_test_pred)} ')
    
    print(f'中值绝对误差为": {median_absolute_error(y_test, y_test_pred)} ')

建模数据+和核心代码下载：1积分