Personalisation of News Recommendation Using Genetic Algorithm

Abstract

News recommendation systems should adapt to individual user preferences to provide news articles of interest to the user. This paper describes personalized news recommendation based on dynamic user profiling, collaborative filtering, location awareness and location sentiments. Dynamic user profiling considers changing user interests. Collaborative filtering approach captures user's behaviour in relationship with other similar users. Providing news about the city that the user is currently stationed will also improve the relevancy of the news. Including news based on the sentiments of the people in the current location is also considered. This paper proposes a news recommendation framework based on four personalisation attributes, namely, user profile, group interest, location and sentiments. The weight age of these attributes is not the same for all users. Hence genetic algorithm is used to identify the weight age of the attributes and personalise it for a particular user. The major issue with the personalized news recommendation system is scalability. This paper addresses the issue using Hbase, a column family database on hadoop framework. Experiments on a collection of sports related news obtained from various news websites demonstrate the efficiency of the proposed approach.