《Python机器学习基础教程》第二章笔记：random_state的作用

男娘i 2023-02-18 14:11 21阅读 0赞

### 《Python机器学习基础教程》第二章笔记：random\_state的作用 ###

*  一、random\_state的作用：固定系数与截距
 *  二、random\_state的取值是对系数排序的结果，random\_state值越小，系数越大

# 一、random\_state的作用：固定系数与截距 #

random\_state的作用在于固定lr.coef\_、lr.intercept\_，保证每次模型的系数、截距一致

不加random\_state时，系数与截距不停的变化：

from sklearn.linear_model import LinearRegression
    import mglearn
    from sklearn.model_selection import train_test_split
    X, y = mglearn.datasets.make_wave(n_samples=60)
    X_train, X_test, y_train, y_test = train_test_split(X, y)
    
    lr = LinearRegression().fit(X_train, y_train)
    print("lr.coef_:", lr.coef_)
    print("lr.intercept_:", lr.intercept_)

第一次：  
![在这里插入图片描述][20200618102908872.png]  
第二次：  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3B5dGhvbl9fcmVwb3J0ZWQ_size_16_color_FFFFFF_t_70]  
循环一下会发现每次系数和截距均不一样：

![在这里插入图片描述][20200618103209763.png]

增加random\_state后：  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3B5dGhvbl9fcmVwb3J0ZWQ_size_16_color_FFFFFF_t_70 1]

# 二、random\_state的取值是对系数排序的结果，random\_state值越小，系数越大 #

第一次取值![在这里插入图片描述][2020061810341131.png]  
系数与截距：  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3B5dGhvbl9fcmVwb3J0ZWQ_size_16_color_FFFFFF_t_70 2]  
第二次取值  
![在这里插入图片描述][20200618103658144.png]  
系数与截距：

![在这里插入图片描述][20200618103709930.png]

第三次取值：  
![在这里插入图片描述][20200618103741924.png]  
系数与截距：

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3B5dGhvbl9fcmVwb3J0ZWQ_size_16_color_FFFFFF_t_70 3]

第四次取值：  
![在这里插入图片描述][20200618103516543.png]

系数与截距：  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3B5dGhvbl9fcmVwb3J0ZWQ_size_16_color_FFFFFF_t_70 4]

现象：系数越来越小，截距看不出来规律！！！  
结论：random\_state取值越小，系数越大；random\_state取值越大，系数越小；两者呈现反比关系

[20200618102908872.png]: https://img-blog.csdnimg.cn/20200618102908872.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3B5dGhvbl9fcmVwb3J0ZWQ_size_16_color_FFFFFF_t_70]: https://img-blog.csdnimg.cn/20200618102933213.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3B5dGhvbl9fcmVwb3J0ZWQ=,size_16,color_FFFFFF,t_70
[20200618103209763.png]: https://img-blog.csdnimg.cn/20200618103209763.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3B5dGhvbl9fcmVwb3J0ZWQ_size_16_color_FFFFFF_t_70 1]: https://img-blog.csdnimg.cn/2020061810331062.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3B5dGhvbl9fcmVwb3J0ZWQ=,size_16,color_FFFFFF,t_70
[2020061810341131.png]: https://img-blog.csdnimg.cn/2020061810341131.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3B5dGhvbl9fcmVwb3J0ZWQ_size_16_color_FFFFFF_t_70 2]: https://img-blog.csdnimg.cn/20200618103433132.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3B5dGhvbl9fcmVwb3J0ZWQ=,size_16,color_FFFFFF,t_70
[20200618103658144.png]: https://img-blog.csdnimg.cn/20200618103658144.png
[20200618103709930.png]: https://img-blog.csdnimg.cn/20200618103709930.png
[20200618103741924.png]: https://img-blog.csdnimg.cn/20200618103741924.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3B5dGhvbl9fcmVwb3J0ZWQ_size_16_color_FFFFFF_t_70 3]: https://img-blog.csdnimg.cn/20200618103755864.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3B5dGhvbl9fcmVwb3J0ZWQ=,size_16,color_FFFFFF,t_70
[20200618103516543.png]: https://img-blog.csdnimg.cn/20200618103516543.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3B5dGhvbl9fcmVwb3J0ZWQ_size_16_color_FFFFFF_t_70 4]: https://img-blog.csdnimg.cn/20200618103504900.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3B5dGhvbl9fcmVwb3J0ZWQ=,size_16,color_FFFFFF,t_70