data.py 文件源码

def load_data(self):
        logging.info('loading the dataset from %s' %self.data_home)
        train_file = os.path.join(self.data_home, 'user_info.train.gz')
        dev_file = os.path.join(self.data_home, 'user_info.dev.gz')
        test_file = os.path.join(self.data_home, 'user_info.test.gz')

        df_train = pd.read_csv(train_file, delimiter='\t', encoding=self.encoding, names=['user', 'lat', 'lon', 'text'], quoting=csv.QUOTE_NONE, error_bad_lines=False)
        df_dev = pd.read_csv(dev_file, delimiter='\t', encoding=self.encoding, names=['user', 'lat', 'lon', 'text'], quoting=csv.QUOTE_NONE, error_bad_lines=False)
        df_test = pd.read_csv(test_file, delimiter='\t', encoding=self.encoding, names=['user', 'lat', 'lon', 'text'], quoting=csv.QUOTE_NONE, error_bad_lines=False)
        df_train.dropna(inplace=True)
        df_dev.dropna(inplace=True)
        df_test.dropna(inplace=True)
        df_train['user'] = df_train['user'].apply(lambda x: str(x).lower())
        df_train.drop_duplicates(['user'], inplace=True, keep='last')
        df_train.set_index(['user'], drop=True, append=False, inplace=True)
        df_train.sort_index(inplace=True)
        df_dev['user'] = df_dev['user'].apply(lambda x: str(x).lower())
        df_dev.drop_duplicates(['user'], inplace=True, keep='last')
        df_dev.set_index(['user'], drop=True, append=False, inplace=True)
        df_dev.sort_index(inplace=True)
        df_test['user'] = df_test['user'].apply(lambda x: str(x).lower())
        df_test.drop_duplicates(['user'], inplace=True, keep='last')
        df_test.set_index(['user'], drop=True, append=False, inplace=True)
        df_test.sort_index(inplace=True)
        self.df_train = df_train
        self.df_dev = df_dev
        self.df_test = df_test