MODIFICATION OF BI-ECLAT ALGORITHM FOR ASSOCIATION RULES MINING IN LARGE TRANSACTIONAL DATA

• Main Author: Erfariani, Nabilah
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/78119
• Institution: Institut Teknologi Bandung

This research project focuses on optimizing algorithms used in the Association Rules Mining method when applied to large transactional data. Given the suboptimal performance of traditional algorithms in handling large transactional datasets, this study aims to develop and implement optimizations for the Bi-Eclat algorithm, referred to as the 'Modified Bi-Eclat Algorithm.' To achieve this objective, the research adopts and modifies several methods such as Priority Queue, down-closure property, and pruning based on support and confidence in the context of Association Rules Mining. By applying these properties, the algorithm is optimized to eliminate subsets that do not meet the criteria for association rules and to process strong association rules earlier in the computation, with the expectation of significantly reducing computational load. Test results indicate that the modified Bi-Eclat Algorithm reduces computation time compared to the traditional Bi-Eclat algorithm while maintaining the same rule quality. Additionally, complexity analysis of the algorithms demonstrates that Modified Bi-Eclat has lower complexity than Bi-Eclat. In conclusion, this research successfully demonstrates that the use of Priority Queue, down-closure property, and pruning based on support and confidence in the context of Association Rules Mining can enhance the efficiency of the Bi-Eclat algorithm in processing large-sized datasets while preserving rule quality. This contributes significantly to improving the performance of Association Rules Mining in Market Basket Analysis on large transactional data.