An enhanced latent semantic indexing with term frequency-inverse document frequency variant for software traceability

Abstract

This paper proposes an improved Latent Semantic Indexing (LSI) with Term Frequency-Inverse Document Frequency (TFIDF) variant for software traceability. The main advantage of the method is the simplicity and accuracy of the algorithm such that it can produce high precision and recall. The accuracy of tracing precise links between software artefacts is achieved by feeding the LSI with TF-IFDF variant that halves the TF and IDF values, thus increases the probability of yielding precise links. To test the accuracy of the proposed method, two case studies were evaluated, namely Mushroom Management System (MMS) and Robotic Wheelchair System (RWS). The superiority of the proposed method over a conventional LSI is confirmed by 409.09%, 900.00%, and 620.00% improvements for precision, recall, and harmonic mean scores, respectively, in MMS case study, at cosine threshold of 0.94. It is anticipated that the method could be beneficial in the design of a practical and accurate system for retrieving precise software traceability links.