Hybrid SVM-TLBO model for improved frequent itemset classification: a performance comparison
Seenam Bee1 and Animesh Kumar Dubey1
Corresponding Author : Seenam Bee
Recieved : 25-December-2024; Revised : 10-March-2025; Accepted : 14-March-2025
Abstract
Frequent itemset classification is a critical task in data mining, enabling efficient discovery of patterns in large transactional and categorical datasets. In this work, a hybrid support vector machine–teaching learning-based optimization (SVM–TLBO) model is proposed to enhance classification performance. The pipeline consists of four stages: preprocessing and transformation of transactional datasets, frequent itemset mining with Apriori, construction of discriminative feature vectors, and optimization of SVM parameters through TLBO. The model was validated on benchmark datasets, namely T10I4D100K and Mushroom, obtained from the UCI repository. Experimental evaluation using precision, recall, F1-score, accuracy, and AUC-ROC metrics demonstrated that the proposed hybrid approach outperformed baseline SVM. Notable improvements were observed in recall and AUC values, highlighting the ability of the SVM–TLBO model to reduce false negatives and improve decision boundary discrimination. The results confirm the effectiveness, scalability, and robustness of the hybrid model for frequent itemset classification in both large-scale transactional and categorical domains.
Keywords
Frequent itemset mining, Support vector machine (SVM), Teaching learning-based optimization (TLBO), SVM-TLBO.
Cite this article
Bee S, Dubey AK. Hybrid SVM-TLBO model for improved frequent itemset classification: a performance comparison. International Journal of Advanced Computer Research. 2025;15(70):1-6. DOI : 10.19101/IJACR.2025.1570010