Improving Travel Recommendation Accuracy using Fusion of Machine Learning Techniques

Abstract

Data recommendation is a multi-processing task, which requires a set of operations ranging from input pre-processing, clustering, similarity analysis to prediction of data-trend. The trend prediction unit performs different kinds of calculations like classification, pattern analysis, error correction, etc. In order to develop a highly accurate recommendation system, the software designers have to accurately measure the performance of each of these algorithms and then identify the best fitting method that optimizes the accuracy of prediction. In our research, we many different combinations of these algorithms, and came up with a list of algorithms which may be best suited for obtaining higher level of prediction accuracy. These algorithms were later implemented, interfaced and integrated with each other in order to evaluate their performance. The performance of the proposed hybrid recommender is found to be 20% more effective than the existing state-of-the art recommenders. This paper also suggests improvement in the proposed system in order to further enhance the performance in terms of speed of operation and accuracy of recommendation.