State-of-the-Art in Empirical Validation of Software Metrics for Fault Proneness Prediction: Systematic Review

Isong, Bassey; Ekabua, Obeten O.

doi:10.5121/ijcses.2015.6601

Cited by 3 publications

(1 citation statement)

References 42 publications

(233 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…So, they obtained an understanding of the degree of accuracy necessary for testing in practical scenarios [15][16][17][18][19]. Bassey et al [20] demonstrated a strong correlation between complexity, coupling, size, and failure prediction. Rath et al, 2022[21] examined the influence of feature selection and ensemble techniques on the Software Reliability Prediction (SRP).…”

Section: Related Workmentioning

confidence: 99%

An Optimised Hybrid ELSVM-BRO Model for Predicting Software Reliability

Rath,

Sahu,

Das

et al. 2024

Preprint

View full text Add to dashboard Cite

In recent years, there has been a rise in strict environmental and safety regulations, resulting in the implementation of extra protocols dictating the functioning and state of software to effectively adhere to safety standards. As a result, the importance of timely, effective, and accurate maintenance procedures has grown significantly. Proper utilization of data has the potential to boost efficiency, reinforce safety measures, lower operational expenses, protect assets, enhance workforce productivity and advance environmental preservation efforts within the software industry. This research aims to devise a novel methodology capable of synchronizing data gathered from multiple sources and constructing a scalable framework to identify early indications of software malfunction. The proposed approach, explored in this study, integrates various Hybrid Extreme Learning Machine (ELM) and Support Vector Machine (SVM) with Binary Rao optimization (JAYA algorithm) techniques (ELSVM-BRO), directly evaluating time series data from the dataset. Pre-processing stages encompass data smoothing, filtering, outlier mitigation, and segmentation, followed by feature extraction for classification purposes. In the given context, a unique model is proposed. This model is a combination of Hybrid Extreme Learning and Support Vector Model, and it’s based on Binary Rao (BR) i.e., also known as Jaya Optimization. The primary purpose of this model is to evaluate the condition of a software system, specifically determining whether it’s faulty or healthy. Comparison with K-Nearest Neighbours (KNN), SVM, and Naïve Bayes (NB) and Random Forest (RF) classifiers using 10 datasets reveals that the ELSVM-BRO model attains superior balanced accuracy levels. The study suggests that amalgamating these algorithms enhances predictive reliability, particularly when applied to datasets of varying sizes.

show abstract