任务1：模型训练与预测

步骤1：导入LightGBM库
步骤2：使用LGBMClassifier对iris进行训练。
步骤3：将预测的模型对iris进行预测。

1.1 导包

import numpy as np
import pandas as pd
import lightgbm as lgb
import json
from sklearn import datasets

# 读取数据
iris = datasets.load_iris()  #载入数据集
# iris

1.2 构建数据集

from sklearn.model_selection import train_test_split

# 将原始数据划分为训练，测试，验证集
train_data_all,test_data,train_y_all,test_y = \
                train_test_split(iris.data, iris.target,test_size=0.2,random_state=1,shuffle=True,stratify=iris.target)
train_data,val_data,train_y,val_y = \
                train_test_split(train_data_all, train_y_all,test_size=0.2,random_state=1,shuffle=True,stratify=train_y_all)

1.3 训练模型

# 构建数据集
lgb_train = lgb.Dataset(train_data,label=train_y)
lgb_val = lgb.Dataset(val_data,label=val_y)

params={
    'learning_rate':0.1,
    'lambda_l1':0.1,
    'lambda_l2':0.2,
    'max_depth':4,
    'objective':'multiclass',
    'num_class':3,  #lightgbm.basic.LightGBMError: b‘Number of classes should be specified and greater than 1 for multiclass training‘
}

# 训练模型
clf=lgb.train(params,lgb_train,valid_sets=[lgb_val])

1.4 模型预测

# 进行预测
from sklearn.metrics import roc_auc_score,accuracy_score
y_pred=clf.predict(test_data)
# y_pred=[list(x).index(max(x)) for x in y_pred]
# print(y_pred)
# print(accuracy_score(y_test,y_pred))

print(y_pred)

[[1.92028726e-04 2.15779127e-03 9.97650180e-01]
 [9.96978554e-01 2.88775364e-03 1.33691922e-04]
 [3.92278727e-02 9.52641287e-01 8.13084015e-03]
 [9.98583095e-01 1.28299744e-03 1.33907086e-04]
 [9.74383148e-01 2.48541945e-02 7.62657557e-04]
 [9.98583095e-01 1.28299744e-03 1.33907086e-04]
 [9.01203243e-04 9.10251290e-04 9.98188545e-01]
 [1.92073483e-04 1.92521742e-03 9.97882709e-01]
 [1.36080663e-03 3.39086128e-02 9.64730581e-01]
 [2.13068261e-03 9.49329566e-01 4.85397510e-02]
 [9.73722014e-01 2.55158454e-02 7.62140083e-04]
 [1.55923053e-03 9.95022567e-01 3.41820246e-03]
 [1.92028726e-04 2.15779127e-03 9.97650180e-01]
 [1.80442736e-03 9.93343715e-01 4.85185800e-03]
 [6.24614547e-04 6.56567666e-03 9.92809709e-01]
 [9.40622432e-01 5.80193844e-02 1.35818314e-03]
 [1.92028726e-04 2.15779127e-03 9.97650180e-01]
 [1.82435627e-03 9.93326817e-01 4.84882643e-03]
 [3.92278727e-02 9.52641287e-01 8.13084015e-03]
 [7.48666460e-03 1.45081791e-01 8.47431544e-01]
 [1.37496776e-03 9.98119028e-01 5.06004117e-04]
 [2.26916234e-03 9.96832784e-01 8.98053599e-04]
 [9.98583095e-01 1.28299744e-03 1.33907086e-04]
 [9.94373850e-01 5.49280693e-03 1.33342638e-04]
 [1.92052404e-04 2.03475217e-03 9.97773195e-01]
 [3.38867037e-03 9.91077011e-01 5.53431846e-03]
 [9.73722014e-01 2.55158454e-02 7.62140083e-04]
 [9.96265410e-01 3.60099334e-03 1.33596291e-04]
 [1.59393780e-02 9.81909021e-01 2.15160091e-03]
 [1.39970367e-03 9.93759937e-01 4.84035958e-03]]

任务2：模型的保存与加载

步骤1：将任务1训练得到的模型，使用pickle进行保存。
步骤2：将任务1训练得到的模型，使用json进行保存。
步骤3：加载步骤1和步骤2的模型，并进行预测。

2.1 使用pickle保存模型

import pickle
pickle.dump(clf, open('model', 'wb'))

2.2 使用json保存模型

import json
model_json = clf.dump_model()
json.dump(model_json,open('model.json','w'))

2.3 加载模型，进行预测

clf = pickle.load(open('model','rb'))

y_pred=clf.predict(test_data)

y_pred

2.4 另外一种方法

# 保存
clf.save_model('model.txt')
# 加载
clf = lgb.Booster(model_file='model.txt') # 注意这里指定model_file