TCL

Introduction

To further improve the recommendation performance, this paper proposes an effective Web API recommendation approach via exploring textual and structural semantics with Trustworthy contrastive learning, named TCL.

Environment Requirment

This code has been tested running undeer Python 3.9.0 The Required packages are as follows:

• torch == 2.0.0+cu118
• numpy == 1.24.1
• seaborn == 0.13.2
• transformers ==4.37.2
• wheel ==0.41.2

Example to run TCL

• Commandpython train_TCL.py
• Train log:

开始训练
存在训练数据，正在加载
存在测试数据，正在加载
10%|█ | 500/5000 [00:09<07:29, 10.02it/s]

NOTE : the duration of training and testing depends on the running environment.
Train environment is on CPU AMD R5 5600x GPU RTX4060ti.

File Introduction

1. model.py

This file contains the code of TCL.

2. sanfm.py

This file contains the code of sanfm.

3. utils.py

This file contains the founction used in the item.

4. train_TCL.py

This file is the model training file.

5. dataset.py

This file contains the dataset loading code.

6. you need to download uncased-bert to the root of this item.
7. data is available in here