An Algorithm for Predicting Frequent Patterns over Data Streams Based on Associated Matrix

Abstract

With the wide application of data mining, many data mining applications need to use past and current data to predict the future state of the data. In view of this situation, we propose a new method, namely AMFP-Stream, for predicting frequent patterns over data streams efficiently and effectively. AMFP-Stream algorithm can predict those frequent item sets that have high potential to become frequent in the subsequent time windows to meet users' needs. Firstly, the algorithm converts the data to 0-1 matrix. Then it will update the associated matrix by tailoring the matrix and bitting operations, from which frequent item sets can be mined as well. Finally, it will predict possible frequent item sets that may appear in the windows next time by using the current data. Experimental results show that AMFP-Stream algorithm can predict the frequent item sets in different experimental conditions, therefore, the algorithm is feasible.