Personalized News Filtering and Recommendation System Using (X2SB-KNN)

In this work a study of personalized news filtering and recommendation systems is presented. An advance K-NN algorithm and its applicability in solving recommendation problems is proposed a Chi-square statistics based (X2SB) version of the K-NN algorithm is proposed. The new X2SB-KNN algorithm can reduce run-time and increases execution speeds through the use of critical X2 value. The recommendation system can overcome scalability problem through Real-Time pattern discovery and online pattern matching. It can also alleviate information overloading and computational complexity problems common with many existing recommendation algorithms. A novel feature selection technique called Fuzzy expert based (FEB) feature selection technique is also proposed this method is used at data pre-processing stage to select the best feature for the classification and recommendation system. An In-house Java program was developed to implement the (X2SB-KNN) classifier on an experimental website. Performance comparison between the proposed system the Euclidean distance K-NN and Naïve Bayesian methods shows that the (X2SB-KNN) classifier can outperform the other methods studied.