机器学习模型保存与加载
id_to_cat是类别编号与类别名称的映射字典:
{0: '出费原因查询', 1: '费用未到账', 2: '账单核实'}
X_test是同训练集一样预处理得到的特征
以分类模型预测作为样例,其中模型保存加载如下:
一、pickle形式
1、保存为pickle
import pickle
# 保存Model(注:save文件夹要预先建立,否则会报错)
with open('/order_analysis/model/svc.pickle', 'wb') as f:
pickle.dump(model, f)
2、读取pickle
# 读取Model
with open('/order_analysis/model/svc.pickle', 'rb') as f:
model_load = pickle.load(f)
# 测试读取后的Model
print([id_to_cat[i] for i in model_load.predict(X_test[2:7])])
二、joblib形式
1、保存pkl
from sklearn.externals import joblib # jbolib模块
joblib.dump(model_load, '/order_analysis/model/svc.pkl')
2、读取pkl
# 读取Model
model_load = joblib.load('/order_analysis/model/svc.pkl')
# 测试读取后的Model
# X_test 需要以训练集同样的方式进行处理
print([id_to_cat[i] for i in model_load.predict(X_test[2:7])])