An Improved Association Rule Mining Algorithm Based on Ant Lion Optimizer Algorithm and FP-Growth

Abstract

Discovering knowledge from the amount of data plays an important role in the era of big data and FP-Growth algorithm is one of the most successful methods for learning association rules. Though the FP-Growth algorithm only needs scan two times, it has a poor efficiency for large datasets. There are already efforts have been made to solve the problem by using some Meta-heuristic optimization algorithms, such as particle swarm optimization algorithm (PSO), immune algorithms etc, which outperform the traditional FP-Growth algorithm and shows strong performance. However, PSO is easy to trap in the local optimums. A novel algorithm ant lion optimizer (ALO) was proposed and with the advantages of global optimization, good robustness, and high convergence accuracy, which was applied to many engineering fields like antenna array synthesis, integrated process planning, scheduling and so on. In the paper, a novel association rule extraction algorithm is put forward based on the ant lion optimization algorithm. A new fitness schema based on confidence and support has been used in this approach, which avoids part of unnecessary searching processes of the FP-Growth algorithm and leads the method of searching the optimization solution more effectively. In order to evaluate the effectiveness of our approach, experiments on various datasets are carried out and experimental results are compared with some other classical meta-heuristic algorithms, experimental results testify the performance of the proposed method.