Title: Tourist volume prediction using data mining techniques and change point detection for Sri Lanka
Authors: B.R.P.M. Basnayake; N.V. Chandrasekara
Addresses: Department of Statistics and Computer Science, Faculty of Science, University of Kelaniya, Kelaniya, 11600, Sri Lanka ' Department of Statistics and Computer Science, Faculty of Science, University of Kelaniya, Kelaniya, 11600, Sri Lanka
Abstract: Sri Lanka is the heart of the Indian Ocean which attracts tourists around the world. This study investigates the behaviour of tourist arrivals in Sri Lanka using data mining techniques and change point analysis accompanied by the main objective of forecasting the tourist volume. Time delay neural network (TDNN), feedforward neural network (FFNN) with Levenberg-Marquardt (LM) and scaled conjugate gradient (SCG) algorithms were applied in forecasting whereas two Windows (WA and WB) were identified with the change point detection. For the entire study period, FFNN with LM algorithm illustrates better performance. A change point was detected in October 2011 in the data. For WA, there was no better-performed model due to fluctuations in tourist arrivals because of terrorist activities. In WB, the outperformed model was obtained from the FFNN with LM algorithm. This study will assist tourism-related industries in their future plans and support in developing infrastructure and economy.
Keywords: change point analysis; data mining techniques; forecasting; neural network; Sri Lanka; tourism.
International Journal of Data Science, 2021 Vol.6 No.3, pp.205 - 222
Received: 02 Mar 2021
Accepted: 31 Aug 2021
Published online: 24 Feb 2022 *