An Intelligent Framework based on CNN and Neutrosophic Technique for Detecting Code Smells and Ranking Code

A code smell is an overt sign that typically points to a more serious issue with the system. The source code has a quality issue, as indicated by the code smell. A software system that has too many smells is challenging to create and maintain. It has taken a lot of work to use metric-based and heuristic-based methods to find smells in Java source code. Recently, machine learning techniques have been used in the context of odor detection; nonetheless, existing techniques are thought to be far from mature. However, researchers and software developers are addressing the aforementioned problems by looking at some of the difficulties associated with recognizing odors using a deep learning-based approach. The big challenge for any company is to choose the best code for their organization in the development process according to their requirements. The main issue in this problem is how to choose the code between the available codes especially when there are many numbers of different types of code smells. This paper presents a novel framework for detecting different types of code smells using CNN. The second main stage in the framework is to rank the available code according to the requirements using multi criteria decision making algorithms. The CNN has achieved accuracy near .955 in the detection process. In the second stage the neutrosophic technique has achieved results better than the other multi criteria decision making algorithms.