An optimal formulation of feature weight allocation for CBR using machine learning techniques

Case based reasoning (CBR) is frequently used for data classification problems, it can be considered as similarity based reasoning but equal importance are assigned to every attribute in the dataset. By identifying the features which are more important in the process of classification we can have better accuracy of CBR system. This paper proposes use of ranked attribute selection on the basis of their relevance. This attribute ranking is done by assigning different weights to different features. The results obtained after conducting experiments also indicated great improvement in the overall accuracy while classifying similar cases in CBR systems. The results are compared by using three famous ranking methods on three different datasets. The obtained results show that the proposed method is effective in terms of ranking the relevant features as compare to irrelevant features.