sklearn随机森林模型：ValueError: Unknown label type: ‘unknown‘

Posted 2022-03-22 Data+Science+Insight

sklearn随机森林模型：ValueError: Unknown label type: 'unknown'

目录

sklearn随机森林模型：ValueError: Unknown label type: 'unknown'

问题：

解决：

完整错误：

---

问题：

分类的标签数据被客户错误地标注为了浮点型；

让学生去做出现了问题，有问题的程序如下；

def random_forest_selection(df):
    
    clf = RandomForestClassifier(n_estimators=100,\\
                                 random_state=42,\\
                                 class_weight = 'balanced',\\
                                 min_samples_leaf = 4,\\
                                 min_samples_split = 4,\\
                                 max_depth = 5,
                                 n_jobs = -1,)

    clf_X = df.drop(['label'], axis=1)
    clf_y = df.label
    
    model = clf.fit(clf_X,clf_y)
    #feat_importances = pd.DataFrame(model.feature_importances_, index=clf_X.columns, columns=["importance"])
    feature_importances = pd.DataFrame('feature': clf_X.columns, 'importance': model.feature_importances_)
    feature_importances.sort_values(by='importance', ascending=False, inplace=True)
    #feat_importances[:10].plot(figsize(12,5),kind='bar',)
    plot_feature_importances(feature_importances, n = 10, threshold = 0.95)
    most = min(50,df.shape[1]//5)
    return feature_importances.feature[0:100].values


my_important = random_forest_selection(df_in)
my_important

解决：

初始数据类型为float64

转换时，正确的书写格式为Int64

这样转换可以转换过去，但是在进入模型的时候会发生问题，

所以，终极的正确的的处理方式是：

缺失值填充后进行数据格式转换；且使用小写int64

df_in = df_in.fillna(0)
df_in['label'] = df_in['label'].astype('int64')

df_in = df_origin
df_in['label'] = df_in['label'].astype("int")
#'Int64'
#df_in['label'] = df_in['label'].astype("Int64")

df_in['label'].value_counts()
# df_in['label'].describe()

完整错误：

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-41-511d3bf43def> in <module>
----> 1 my_important = random_forest_selection(df_in)
      2 my_important

<ipython-input-40-4490edd7b355> in random_forest_selection(df)
     12     clf_y = df.label
     13 
---> 14     model = clf.fit(clf_X,clf_y)
     15     #feat_importances = pd.DataFrame(model.feature_importances_, index=clf_X.columns, columns=["importance"])
     16     feature_importances = pd.DataFrame('feature': clf_X.columns, 'importance': model.feature_importances_)

D:\\anaconda\\lib\\site-packages\\sklearn\\ensemble\\_forest.py in fit(self, X, y, sample_weight)
    329         self.n_outputs_ = y.shape[1]
    330 
--> 331         y, expanded_class_weight = self._validate_y_class_weight(y)
    332 
    333         if getattr(y, "dtype", None) != DOUBLE or not y.flags.contiguous:

D:\\anaconda\\lib\\site-packages\\sklearn\\ensemble\\_forest.py in _validate_y_class_weight(self, y)
    557 
    558     def _validate_y_class_weight(self, y):
--> 559         check_classification_targets(y)
    560 
    561         y = np.copy(y)

D:\\anaconda\\lib\\site-packages\\sklearn\\utils\\multiclass.py in check_classification_targets(y)
    181     if y_type not in ['binary', 'multiclass', 'multiclass-multioutput',
    182                       'multilabel-indicator', 'multilabel-sequences']:
--> 183         raise ValueError("Unknown label type: %r" % y_type)
    184 
    185 

ValueError: Unknown label type: 'unknown'