Multi-Label Code Smell Detection with Hybrid Model based on Deep Learning
Multi-Label Code Smell Detection with Hybrid
Model based on Deep Learning
Yichen Li
, Xiaofang Zhang
*
School of Computer Science and Technology Soochow University
SEKE 2022
2
Introduction
•
Code Smells indicate problems related to aspects of code quality such
as understandability and modifiability, and imply the possibility of
refactoring.
•
Many approaches have been proposed to detect code smells
but
they
lose much information which helps recognize each code smell more
efficiently and pay little attention to multi-label code smell detection.
•
In this paper, we propose a hybrid model with multi-level code
representation to further optimize the code smell detection. Better
results have been achieved on 100 high-quality Java projects from
Github.
Method-HMML
3
Method-HMML
4
Dataset
•
We choose nine code smells at the method level for our
experiment and combine their labels into a multi- label dataset.
5
Evaluation -
Results
•
RQ1:How does our HMML method perform compared to other baselines?
6
•
We can regard that our HMML method does a good job in the
multi-label code smell detection.
Evaluation -
Results
•
RQ
2
: How does multi-label code smell detection perform compared
with
single
code smell detection?
7
•
our multi-label code smell detection not only performs well in each code smell
detection but can find out all code smells by one model at the same time.
Evaluation -
Results
•
RQ3: What impact does each of our main components have in our model?
8
•
Each part of the model plays its own advantages as expected
and captures the corresponding code smell features
Conclusion
•
We propose a hybrid model that extracts the multi-level code representation
information and separately applies the appropriate deep learning neural
network.
•
We are the first to carry out the multi-label code smell detection based on the
deep learning method and achieve a good result.
•
We modify many other approaches to fit into multi-label classification tasks
and conduct extensive experiments to find the maximum capacity and best
configuration.
9
https://github.com/liyichen1234/HMML
Thanks!
Q&A
Code smells indicate code quality problems and the need for refactoring. This paper introduces a hybrid model for multi-label code smell detection using deep learning, achieving better results on Java projects from Github. The model extracts multi-level code representation and applies deep learning neural networks effectively. Experimental results show the model's good performance compared to baselines and its ability to detect multiple code smells simultaneously.
Download Presentation
Please find below an Image/Link to download the presentation.
The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author.If you encounter any issues during the download, it is possible that the publisher has removed the file from their server.
You are allowed to download the files provided on this website for personal or commercial use, subject to the condition that they are used lawfully. All files are the property of their respective owners.
The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author.
E N D
Presentation Transcript
Multi-Label Code Smell Detection with Hybrid Model based on Deep Learning Yichen Li, Xiaofang Zhang* School of Computer Science and Technology Soochow University SEKE 2022 SOOCHOW UNIVERSITY
Introduction Code Smells indicate problems related to aspects of code quality such as understandability and modifiability, and imply the possibility of refactoring. Many approaches have been proposed to detect code smells but they lose much information which helps recognize each code smell more efficiently and pay little attention to multi-label code smell detection. In this paper, we propose a hybrid model with multi-level code representation to further optimize the code smell detection. Better results have been achieved on 100 high-quality Java projects from Github. 2
Dataset We choose nine code smells at the method level for our experiment and combine their labels into a multi- label dataset. 5
Evaluation - Results RQ1:How does our HMML method perform compared to other baselines? We can regard that our HMML method does a good job in the multi-label code smell detection. 6
Evaluation - Results RQ2: How does multi-label code smell detection perform compared with single code smell detection? our multi-label code smell detection not only performs well in each code smell detection but can find out all code smells by one model at the same time. 7
Evaluation - Results RQ3: What impact does each of our main components have in our model? Each part of the model plays its own advantages as expected and captures the corresponding code smell features 8
Conclusion We propose a hybrid model that extracts the multi-level code representation information and separately applies the appropriate deep learning neural network. We are the first to carry out the multi-label code smell detection based on the deep learning method and achieve a good result. We modify many other approaches to fit into multi-label classification tasks and conduct extensive experiments to find the maximum capacity and best configuration. https://github.com/liyichen1234/HMML 9
Thanks! Q&A
