Abstract
In this study, we propose a fast algorithm to form product-based fuzzy association rules from large quantitative dataset, which reduces data size and ensures the quality of the obtained results. A method is designed to transform mining of fuzzy association rules to the binary counterpart. It is shown that the final results are not affected by this transformation. Then, an efficient sampling method is developed, where a sample is taken to replace the original large dataset, so the size of the dataset is reduced and the cost of scanning is also decreased. Through the central limit theorem, the size of sample can be set reasonably, so the deviation of support of any fuzzy itemset caused by sampling is limited in a small range with a high probability. Through a series of experiments, we show the advantages of the approach both the speed of the proposed algorithm and its reliability. (C) 2017 Elsevier B.V. All rights reserved.