KNN algorithm is one of the best and the most usable classification algorithms which is used largely in different applications. One of the problems of this algorithm is the equal effect of all attributes in calculating the distance between the new record and the available records in training dataset ,how ever ,may be some of these attributes are less important to the classification and some of these attributes are more important. This results in misleading of classification process and

more »

... g the accuracy of classification algorithm. A major approach to deal with this problem is to weight attributes differently when calculating the distance between two records. In this research we used association rules to weight attributes and suggested new classification algorithm K-Nearest-Neighbor-Based-Association (KNNBA) that improves accuracy of KNN algorithm. We experimentally tested KNNBA accuracy, using the 15 UCI data sets [1], and compared it with other classification algorithms NB, NN, C4.4, NBTREE, VFI, LWL and IBK. The experimental results show that KNNBA outperforms these seven algorithms.