Fuzzy Wavelet Dynamic Neural Network Model for Modeling the Number of Tourist Visits to West Nusatenggara Province

Abstract

This research aims to model tourist visit data to West Nusa Tenggara Province (NTB) using a hybrid model, combining the dynamic neural network method as the core model with the wavelet method and fuzzy inference as tools to optimize the model. The model developed in this research is referred to as the Fuzzy Wavelet Dynamic Neural Networks (FW-DNN) Model. The FW-DNN model is a feed-forward dynamic neural network model that utilizes the Wavelet B-spline function as its activation function and TSK (Takagi-Sugeno-Kang) fuzzy inference as the method for information aggregation. The modeling results on both in-sample and out-sample data show that the proposed FW-DNN model is capable of representing the patterns in tourist visit data to NTB quite effectively. Similar results were also observed in the patterns of data for both domestic and international tourist visit numbers. Based on the root of mean square error (RMSE) indicator, the performance of the developed FW-DNN model for aggregated tourist visit data is 95185.09 for in-sample data and 22615.54 for out-sample data. Partial performance analysis of the FW-DNN model for international tourist visit data shows a value of 39848.94 for in-sample data and 5223.86 for out-sample data. Similarly, the FW-DNN model's performance for domestic tourist visit data is 39848.94 for in-sample data and 5223.86 for out-sample data. Practically, the results of this research can be input for the NTB Provincial government in determining tourism management and/or development policies, especially related to the provision of supporting facilities and infrastructure, or the private sector in an effort to optimize the carrying capacity and/or services for tourists visiting NTB.