Robust Classification Based on Correlations Between Attributes

The existence of noise in the data signiflcantly impacts the accuracy of classiflcation. In this paper, we are concerned with the development of novel classiflcation algorithms that can e-ciently handle noise. To attain this, we recognize an analogy between k nearest neighbors (kNN) classiflcation and user-based collaborative flltering algorithms, as they both flnd a neighborhood of similar past data and process its contents to make a prediction about new data. The recent development of item-based collaborative flltering algorithms, which are based on similarities between items instead of transactions, addresses the sensitivity of user- based methods against noise in recommender systems. For this reason we focus on the item-based paradigm to provide improved robustness, compared to kNN algorithms, against noise for the problem of classiflcation. We propose two new item-based algorithms, which are experimentally evaluated with kNN. Our results show that, in terms of precision, the proposed methods outperform kNN classiflcation by up to 15%, whereas compared to other methods like the C4.5 system, improvement exceeds 30%.