Category Based Location Aware Tourist Place Popularity Prediction and Recommendation System Using Machine Learning Algorithms

Abstract

AbstractTourism contributes majorly in the economic growth of a country. Country like India which is a large market for tourism. Tourism is one of the prime sectors contributing highly to the GDP. Hence, it is necessary to build a smart tourism system which helps in increasing revenue from the tourism industry. In this paper, we have proposed a novel framework for place popularity prediction and recommendation system using machine learning algorithms. A novel recommendation system can overcome cold start problem. Dataset used in a research study has been gathered from popular tourism web sites. In experiment, category of places has been determined using the LDA algorithm. Location of places has been identified using K-means clustering algorithm. Sentiment analysis has been used for popularity rating prediction. Sentiment classification has been done using famous supervised machine learning algorithms, i.e., naive Bayes (NB), decision tree (DT), support vector machine (SVM), random Forest (RF), and performance analysis using several evaluation factors. From research, we could conclude that random forest has given the highest performance in comparison with decision tree, naive Bayes, and SVM classifiers. As a result, popularity-based tourist spot classification using RF had been implemented which has given accuracy of 88.02% for testing data used. On bases of category + location top-N popular places have been recommended to an end user using a combination of LDA, and RF and K-means algorithms used in a layered approach. This hybrid recommendation system is a combination of content-based + popularity-based recommendation systems. Such a recommendation system will give more precise recommendations as compared to only content-based and only popularity-based systems.KeywordsLatent Dirichlet allocationLinear support vector machineMultinomial naive BayesBag of wordsTF-IDFK-meansRandom forestDecision tree