RECOMMENDATION BASED APPROACH USING FCM IN E-COMMERCE SYSTEM

Abstract

Aim of recommendation system is to predict the level of preferences of users towards some items, with the purpose of suggesting ones that they may like, among the sets of items they have not visited yet. In other words the aim of recommendation system is to improve the user experience by suggesting items the user may purchase. The recommendation system targets the personalization aspects of each user. It provides the items in the recommendation list according to particular customers and improve the user experience and for this it exploits the user specific interest and the preferences by his profile. Recommendation system applies in E-Commerce, Web search and the advertising fields. The main aim of recommendation is to increase the sale in ECommerce System. Recent researches have proved that a good recommendation system always increases the sale by good amount of percentage. Recommendation System also suffers from some problems. It depends upon the following factors:  If the user request is not clear and result to multiple interpretations then this can result to the collection the responses for each of those meanings, and then deciding which result we can present to the user.  If the retrieved results are very similar to each other and if one of them is relevant then we have to present all of them, this can present lots of unnecessary data  The profiles, needs, and activities of different users can be very different and that’s why we have to make a system that can response to all type of queries submitted to that. We can solve these kinds of problems by classification and clustering techniques and for this purpose we use the clustering approach to perform the classification of user profile, interests and the items. It performs the partitioning of the data from the large data set to produce the short representation of the recommendation system's behaviour. Fuzzy c-means (FCM) is a data clustering technique in which we divide the dataset into n number of clusters and each data point in the dataset belongs to every cluster in the group of clusters to a certain degree of membership. If a data point lies close (near) to the centre of the cluster, then it will have a high degree of membership to that cluster and another data point which lies away from the centre of the cluster would have the low degree of membership to that cluster. It allows belongingness of one data point to more than one cluster.