K-Means Algorithm and Clustering Technique for A Recommender System

Abstract

Most of netizens are fond of traveling. Though they are interested to travel, majority of them are confused about where to go. Whenever the netizens planed for a travel, they spent hours searching for an interesting place that matches their point of interest. Therefore, there is a need for a recommender system which can recommend several interesting travel venues based on their preferences. Hence, information regarding different users prefers different travel locations are required. In this paper, Google review is used as a reference to divide the users into clusters of similar interests. The data has been divided into six scenarios which consist of the original data, scaled data, Principal Component Analysis (PCA) data, PCA scaled data, MinMax scaled data and Robust scaled data. The distance metric technique is performed in all these scenarios. Minkowski distance metric is used to measure the distance between the data points in the dataset and the elbow method to these scenarios to know the inertia values. This has revealed that the PCA with scaled data has lower distance and inertia value. The K-Means clustering is performed on this PCA-scaled data which divided the data into four different clusters. The results revealed are imperative to build any recommender system based on the users' preference.