Comparative Study of User and Item k-NN for Rating Prediction

Collection of data is providing everybody a great knowledge base in modern world. It supports most of our researches in almost every field of science and technology. The only problem faced by collection of data is that the large repositories of data are not easy to handle both in case of data access and desired data retriev. Data filtering solves this problem up to an extent and provides a better solution when used with prediction techniques. This combination of data filtering and prediction provided the recommendation systems. This paper proposes the recommendation system models based on user ratings. Training and testing models are designed to predict user ratings for new users. The models are designed on the basis of an approach named item k-NN. A comparative analysis is performed for item k-NN and user k-NN.