Quantitative Evaluation of Approximate Frequent Pattern Mining Algorithms

Traditional association mining algorithms use a strict definition of support that requires every item in a frequent itemset to occur in each supporting transaction. In real-life datasets, this limits the recovery of frequent itemset patterns as they are fragmented due to random noise and other errors in the data. Hence, a number of methods have been proposed recently to discover approximate frequent itemsets in the presence of noise. These algorithms use a relaxed definition of support and additional parameters, such as row and column error thresholds to allow some degree of “error” in the discovered patterns. Though these algorithms have shown to be successful in finding the approximate frequent itemsets, a systematic and quantitative evaluation approach to evaluate the patterns they generate has been lacking. In this paper, we propose a comprehensive evaluation framework to compare different approximate frequent pattern mining algorithms. The key idea is to select the optimal parameters for each algorithm on a given dataset and use the itemsets generated with these optimal parameters in order to compare different algorithms. We also proposed simple variations of some of the existing algorithms by introducing an additional postprocessing step. Subsequently, we applied our proposed evaluation framework to a wide variety of synthetic datasets with varying amounts of noise and a real dataset to compare existing and our proposed variations of the approximate pattern mining algorithms. Source code and the datasets used in this study are made publicly available.

Source Codes for all the algorithms implemented in this paper can be downloaded from here: Download source codes
All the eight different synthetic datasets used in this study can be downloaded from here: Download datasets
We only showed results on dataset 6 and dataset 8 in the paper due to space constraints. Results on other datasets can be downloaded from here: Download all results