Efficient Data Preprocessing for Genetic-Fuzzy Mining with MapReduce

Genetic-fuzzy data mining can successfully find out linguistic association rules and appropriate membership functions close to human concepts from quantitative transactions, and thus becomes a promising research field in these years. It repeatedly uses fuzzy frequent 1-itemsets to evaluate fitness values of chromosomes, which is very time-consuming. In this paper, we propose a MapReduce preprocessing approach to efficiently transform given quantitative transaction data into pairs of items and quantity lists to increase the performance of genetic-fuzzy mining. The MapReduce architecture totally fits the conversion due to its characteristics of key-value format. Experimental results also show the effect of the proposed approach.