A Systematic Comparison of Search Algorithms for Topic Modelling—A Study on Duplicate Bug Report Identification

Abstract

Latent Dirichlet Allocation (LDA) has been used to support many software engineering tasks. Previous studies showed that default settings lead to sub-optimal topic modeling with a dramatic impact on the performance of such approaches in terms of precision and recall. For this reason, researchers used search algorithms (e.g., genetic algorithms) to automatically configure topic models in an unsupervised fashion. While previous work showed the ability of individual search algorithms in finding near-optimal configurations, it is not clear to what extent the choice of the meta-heuristic matters for SE tasks. In this paper, we present a systematic comparison of five different meta-heuristics to configure LDA in the context of duplicate bug reports identification. The results show that (1) no master algorithm outperforms the others for all software projects, (2) random search and PSO are the least effective meta-heuristics. Finally, the running time strongly depends on the computational complexity of LDA while the internal complexity of the search algorithms plays a negligible role.

The Slow and the Furious? Performance Antipattern Detection in Cyber-Physical Systems

“…Through manual analysis, we identified 530 (81.11%) commits discussing one or more performancerelated issues. We further expanded the keywords by using textual analysis methods (De Lucia et al, 2014) and topic modeling (Panichella et al, 2013;Panichella, 2019). This resulted in 1640 additional commits, of which 163 commits (9.37%) contained one or more selfadmitted performance-related issues.…”

“…Then, we pre-process these artifacts by tokenizing the commit message, removing stop words, and stemming. First, tokenization aims to extract words in the text and remove nonrelevant characters, such as punctuation marks, special characters, and numbers (Panichella, 2019). As commit messages can contain code snippets, we split compound names (i.e., identifiers) into tokens using camel case and snake case splitting (Panichella et al, 2016).…”

It Takes Two to Tango: Combining Visual and Textual Information for Detecting Duplicate Video-Based Bug Reports

“…Reports. Many research projects have focused on detecting duplicate textual bug reports [28,29,35,36,52,54,55,64,70,72,74,82,84,86,87,89,90,[93][94][95][96][97][99][100][101]107]. Similar to Ta n g o , most of the proposed techniques return a ranked list of duplicate candidates [35,63].…”

A Systematic Study of Duplicate Bug Report Detection

“…This step includes stemming and removal of stop words. Then either the Termby-document matrix or the probabilistic models are generated which are then used to calculate the textual similarities [24]. Term-by-Document matrix includes vocabulary which is also referred to as Terms as rows and the documents as the columns.…”

“…[35] used the LDA and LSI approaches to find how continuously querying the bug report like how it happens in Google Search Engine helps find the duplicate bug reports. The author in [24] in their paper compared five metaheuristics GA, DE, Particle Swarm Optimization, Simulated Annealing and Random Search, to analyze how the LDA works when applied.…”