Recommender Model for Secure Software Engineering using Cosine Similarity Measures

Abstract

One of the essential components of Recommender Systems in Software Engineering is a static analysis that is answerable for producing recommendations for users. There are different techniques for how static analysis is carried out in recommender systems. This paper drafts a technique for the creation of recommendations using Cosine Similarity. Evaluation of such a system is done by using precision, recall, and so-called Dice similarity coefficient. Ground truth evaluations consisted of using experienced software developers for testing the recommendations. Also, statistical T-test has been applied in comparing the means of the two evaluated approaches. These tests point out the significant difference between the two compared sets.