DEVELOPMENT OF ASSOCIATION RULE FRAMEWORK USING PARTICLE SWARM OPTIMIZATION ON FP-GROWTH WEIGHTED FREQUENT ITEMSET MINING

• Main Author: Muhammad, Fadly
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/87708
• Institution: Institut Teknologi Bandung

This research develops an association rule framework based on optimizing the minimum support value, minimum confidence, and applying weights to items using a combination of PSO and WFIM algorithms in the FP-Growth algorithm. This research is designed to overcome the limitations in determining the minimum support value and considering the weight of each item. This research aims to develop an association rule optimization framework for the FP-Growth algorithm that produces relevant and efficient association rules. In an effort to overcome the limitations of the traditional FP-Growth algorithm, this research proposes optimization through three main approaches. First, optimization of FP-Growth algorithm using PSO to automatically determine the optimal minimum support and confidence values. Second, the application of the WFIM approach to improve the relevance of association rules by considering the weight of each item in the transaction. Third, the integration of PSO and WFIM in the developed framework creates a more efficient, flexible, and adaptive approach to dataset characteristics. Experimental results show that the combination of PSO and WFIM in FP-Growth can improve the quality of association rules with more significant lift values than traditional methods. In terms of runtime and memory usage, the optimized framework shows higher efficiency, with reduced execution time and lower memory requirements than traditional FP-Growth. The results concluded that the PSO and WFIM approaches improved the flexibility of the FP-Growth algorithm in handling datasets with complex characteristics. The optimized algorithm successfully extracts more association rules with better quality than the traditional FP-Growth algorithm. It also shows reduced runtime and more efficient memory usage.