Classification with the Use of Association Rules with Local Weighting

Classification is one of the best assets in the field of data mining and machine learning. In classification problems one can learn a model or a function by the use of training data, then he or she can use this function (or model) to classify other data that have never been seen before in the system. One of the classification techniques is to extract association rules from data. This technique tries to find meaningful relationships between the members of a dataset. These relationships can be defined by the use of some rules. Recently, the use of associative rules for the purpose of classification captured a lot of attention among scientists. Researches done in this area, showed the high potential of association rules for classification. This article tries to use local weighting of inputs (data) for extracting better rules and then by pruning of association rules enhance the accuracy of classification act. To estimate the performance of the proposed method, two different set of classification algorithms, classic classification and classification based on modern association rules, were compared to it. The acquired results, showed the high capabilities of proposed method in the field of classification.

Keywords:

Association rules , local weighting of inputs , classification.