python连接数据库(Redis、Hive、MySQL、HDF5)

1. 加载模型

1.1 使用pmml加载模型

from pypmml import Model

model = Model.fromFile("lightgbm.pmml")
model.predict(X_test)

1.2  使用joblib加载模型

from sklearn.externals import joblib
model = joblib.load("model_{}.m".format(str(date)))

2. 保存模型

2.2  使用joblib保存模型

import joblib
joblib.dump(model, "model_{}.m".format(str(date)))

3. 字典数据的加载与保存

3.1 存储字典

直接以pickle格式存储字典，每次使用前加载该文件即可。

channel_cat2num_dict = {}
channel_list = df['source_channel'].unique()
for idx, item in enumerate(channel_list):
    channel_cat2num_dict[item] = idx
joblib.dump(channel_cat2num_dict, '../../models/source_channel_cat2num.pkl')

3.2 加载字典

channel_dict = joblib.load('../../models/source_channel.pkl')
test['source_channel'] = test['source_channel'].map(channel_dict)

或
with open('aa.pkl', 'rb') as f:
    channel_dict = pickle.load(f)