Finding the Best Threshold in Association Rule Mining Using the Frog Leaping Algorithm

In exploring association rules, most of the previous studies have been conducted on the optimization of the efficiency, however, determination of the backup, and safety threshold, has a great effect on the quality of the association rules, and is still of importance. In the traditional algorithms for association rule mining, the two parameters of the confidence, are always determined by the deciding user, using trial and error. Despite the high impact of this task on the performance of the algorithms for association rule mining, these algorithms do not have a good performance in Big Data. Therefore, designing a method to automatically find the best value for these parameters, in Big Data, is a necessity. In this article, a new method for enhancing the calculative efficiency in Big Data to determine the best threshold is suggested. In this method, using the Shuffle Frog Leaping Algorithm (SFLA), first the best fitness for each frog is determined, then the minimum support and confidence are specified. The results show that the SFLA could yield better threshold than genetic and a-priori algorithms. The superiority of the suggested method is that it has the capability of functioning in Big Data, and it is more probable to escape the local optimum trap than the genetic algorithm; moreover, its convergence is quick and its search accuracy is high.