【人工智能项目】深度学习实现金融用户评论分类

任务说明

本数据集包含金融行业用户评论信息，其中有大量长文本，根据评论识别所属金融业务，具体类别如下：

0：消费贷款

1：抵押

2：信用卡

3：债务催收

4：信用报告

5：学生贷款

6：银行账户服务

7：短期小额贷款

8：汇款

9：预付卡

10：其他金融服务

认识数据

import pandas as pd

train_df = pd.read_csv("train.csv")
test_df = pd.read_csv("test.csv")

train_df

test_df

CatBoost分类

# 裸模86.9556，主要是认识一下俄罗斯很强大的学习模块Catboost
import pandas as pd
from catboost import CatBoostClassifier
import numpy as np


df1=pd.read_csv("./train.csv")
y = df1['label']
X_train = df1['text'].astype('str')
X_train=np.array(X_train)

df2=pd.read_csv("./test.csv")
X_test = df2['text'].astype('str')
X_test=np.array(X_test)

print(X_test.shape)

model = CatBoostClassifier(task_type='GPU')
model.fit(X_train, y,text_features=[0])

label=[]
for i in range(X_test.shape[0]):
    label.append(model.predict([X_test[i]])[0])

dataframe = pd.DataFrame({'index':range(X_test.shape[0]),'label':label})
dataframe.to_csv("submisson.csv",index=False,header=False,sep=',')

SimpleTransformer模型训练预测

import warnings
warnings.simplefilter('ignore')

import numpy as np

import pandas as pd

from simpletransformers.classification import ClassificationModel

train = pd.read_csv('./train.csv')
test = pd.read_csv('./test.csv')

train.columns = ['text', 'labels']
train.head()

train_args={
    'sliding_window': True,
    'reprocess_input_data': True,
    'overwrite_output_dir': True,
    'logging_steps': 5,
    'stride': 0.6,
    'max_seq_length': 256,
    "fp16":False,
    'num_train_epochs': 5,
    "train_batch_size":1,
    "gradient_accumulation_steps":32
}

model = ClassificationModel('roberta', 'roberta-base', num_labels=11, args=train_args)

model.train_model(train)

predictions, _ = model.predict([row['text'] for _, row in test.iterrows()])
pd.DataFrame({'ID': test.index, 'labels': predictions}).to_csv('submission_roberta-large.csv', index=False, header=False)

小结

本次用机器学习的方法实现nlp相关的评论分类工作，那下次见！！！