Twin Support Vector Machine with Minkowski Gaussian Kernel Based Performance Anomalies Detection in Software Application During Runtime

In the current technological era, software plays a central role across various industries, and detecting anomalies during runtime has become crucial for organizations operating in domains such as commerce, engineering, and Internet of Things. However, a significant challenge arises due to the non-linearly separable nature of defective and non-defective software, leading to substantial overlap in their feature spaces. As a result, the existing Machine Learning (ML) models fail to capture the complex relationships. To overcome this limitation, a Twin Support Vector Machine with Minkowski - Gaussian Kernel (Twin SVM - MGK) is proposed for efficient software defect prediction on the DApp and Vichalana datasets. The proposed Twin SVM - MGK model employs Minkowski - Gaussian Kernel to map the selected features into a higher-dimensional spaceV, thereby enhancing separability of the defective and non-defective modules, especially in non-linear scenarios where intricate relationships are exhibited. Initially, raw data is preprocessed by replacing missing values with standardization techniques for data enhancement. Then, the preprocessed attributes are fed as input for the Chi-Square based feature selection process to perform feature selection, retaining only the relevant information. Finally, the Twin SVM model effectively detects performance anomalies in software applications by differentiating between defective and non-defective modules. The experimental results demonstrate that the proposed Twin SVM-MGK model achieves an impressive accuracy of 99.23%, significantly outperforming existing approaches such as the Random Forest.