執行

token = Tokenizer(num_words=3000)
token.fit_on_texts(train_text)

出現:

--------------------------------------------------------------------------- AttributeError                            Traceback (most recent call last) <ipython-input-250-1f498ce7daab> in <module>()       1 token = Tokenizer(num_words=3000) ----> 2 token.fit_on_texts(train_text)       3 #if lower:       4 #    train_text = train_text.lower()  C:\ProgramData\Anaconda3\envs\tensorflow\lib\site-packages\keras\preprocessing\text.py in fit_on_texts(self, texts)     186                                                                      self.filters,     187                                                                      self.lower, --> 188                                                                      self.split)     189             for w in seq:     190                 if w in self.word_counts:  C:\ProgramData\Anaconda3\envs\tensorflow\lib\site-packages\keras\preprocessing\text.py in text_to_word_sequence(text, filters, lower, split)      37     """      38     if lower: ---> 39         text = text.lower()      40       41     if sys.version_info < (3,) and isinstance(text, unicode):  AttributeError: 'float' object has no attribute 'lower'

在轉入資料時因有部分欄位的資料為NaN

因此在執行token.fit_on_texts(train_text)時會發生如上的錯誤，

可以在資料匯入後先將NaN的空值轉換為''空白，就可防止上面的錯誤。

df = pd.read_csv("gender-classifier-DFE-791531.csv",delimiter=',', encoding='latin1')
df['description'] = df['description'].fillna('')