document_filter.py 文件源码

def predict(self, ifile, efile, ofile):
        # Load
        columns          = ['documents']
        data             = pd.read_table(ifile, header=None, names=columns)
        documents        = data['documents']

        # Deserialize
        estimator        = pickle.load(open(efile, 'rb'))

        # Predict
        probability      = estimator.predict_proba(documents)
        data['labels']   = estimator.predict(documents)
        data['C1_pr']    = probability[:, 0]
        data['C2_pr']    = probability[:, 1]

        # Save
        columns          = ['labels', 'C1_pr', 'C2_pr', 'documents']
        data.to_csv(
            ofile,
            sep          = '\t',
            columns      = columns,
            index        = False
        )