【bug】‘CustomPCA‘ object has no attribute ‘n_features_in_‘

对流层顶的圆白菜

于 2024-05-05 13:05:00 发布

阅读量427

点赞数 2

文章标签： bug

本文链接：https://blog.csdn.net/m0_45208018/article/details/138464362

版权

文章讲述了在使用pyEOF进行旋转因子正交分解时遇到关于CustomPCA对象属性错误的问题，解决方法是更新CustomPCA类添加n_features_in_属性以适应新版sklearn库。作者提供了修改后的代码片段作为参考。

摘要由CSDN通过智能技术生成

问题

在使用pyEOF包（load sample data — pyEOF 0.0.0 documentation）做旋转因子正交分解时出错（运行官方的示例代码时就出错了）

'CustomPCA' object has no attribute 'n_features_in_'

解决

额，好像是因为新的sklearn库将n_features_废弃了，用的是n_features_in_，给CustomPCA类加个n_features_in_属性就好了
我是这样改的，改了可以运行了：

/* pyEOF.py文件 71-75行替换为*/


if pca_type=="varimax":
    skpca = CustomPCA(n_components=n_components, 
                        rotation='varimax',n_features_in_ = data_dropna.shape[1]) #zb data_dropna (time,current features)
    skpca.fit(data_dropna)
    self._EOFs = np.full([n_components, self._space], np.nan)

/* custom_pca.py文件 60-83行替换为*/

def __init__(self, rotation=None, feature_selection='all', n_features_in_=10086, **kws): #zb 240505
    '''
    rotation: string, name of the rotation method.
                'varimax': varimax rotation
                None: no rotation by by default

    feature_selection: string, Features selection method
                        'all': All features are selected, means principal
                                components are output of tranformation
                        'significant': Only features with significant weights
                                        and communalities are selected
                        'surrogate': Only features with highest loading
                                    with principal components are used
                        'summated scales': Summated scales consturcted as
                                            sum of features having significant
                                            loadings on principal components,
                                            not implemented yet.
    '''

    super().__init__(**kws)
    self.rotation = rotation
    self.feature_selection = feature_selection
    # self.n_features_in_ = 10086 #zb 240505 这个初始数字是不是没用 算了做成参数吧
    self.n_features_in_ = n_features_in_