HYBRID MODEL, NEURAL NETWORKS, SUPPORT VECTOR MACHINE, K-NEAREST NEIGHBOR, AND ARIMA MODELS FOR FORECASTING TOURIST ARRIVALS

An autoregressive integrated moving average (ARIMA) model has been succeed for forecasting in various field. This model have disadvantages in handling the non-linear pattern. Artificial Neural Networks (ANN), Support Vector Machine (SVM) and K-Nearest Neighbor (k-NN) models can be considered to handle non-linear pattern. Neural network, SVM and k-NN models have also succeed for forecasting in various fields and these models yield mixed results of performance. In this paper, we propose a hybrid model combining ARIMA and Artificial Neural Networks model with optimum number of neuron in input layer, optimum number of neuron in hidden layer, optimum of activation function for forecasting tourist arrivals. The forecasting accuracies of the models are compared based on tourist arrivals time series data. The proposed hybrid model yield better forecasting accuracies results compared to ARIMA, K-Nearest Neighbor, neural network and Support Vector Machine with various kernel