Archives

User based Collaborative Filtering based on Minkowski Distance Measure to Support Small Medium Enterprise Ecommerce Company


Eka Angga
Abstract

Collaborative filtering as part of data mining and information retrieval research domain has been used in industry especially in ecommerce. Making recommendation about product, movie or music into customers become strategic planning to increase profit. There is need a high-performance technique and algorithm to produce recommendation that match what user need. Basic User Based Collaborative Filtering technique uses cosine similarity to create recommendation. It becomes the weakness caused by the nature of cosine similarity when tries to deal with sparse data. In this paper, we purpose a modified User Based Collaborative Filltering by changing the similarity measure. Objective of this paper is to create recommendation based on user’s similarity by using minkowski distance measure and evaluate the results. Performance measurement approach follow the common evaluation scheme in information retrieval such as Accuracy, Precision, Recall and Fold Validation also involved. Experiment result shows that minkowski distance measure get better result than cosine similarity measure.

Volume 11 | 07-Special Issue

Pages: 646-652