Hybrid Recommendation Mechanisms for Tourism Personalization: A Computational Study in the Indonesian Context

Abstract

The tourism industry in Indonesia has experienced rapid growth, generating large volumes of data on tourist preferences, destinations, travel costs, and durations. Traditional travel recommendations, provided through agencies or static guidebooks, often rely on generalized assumptions and fail to adapt to individual user preferences. Consequently, tourists frequently receive generic suggestions that may not align with their specific needs, resulting in suboptimal travel experiences. To address these limitations, this study proposes a hybrid learning-based travel recommendation system that leverages Indonesian tourism datasets to deliver personalized suggestions. The system employs multiple Classification and Regression Trees (CART)-based models, including Linear Logistic Regression (LLR), Support Vector Machine (SVM), and Random Forest (RF), for comparative analysis. The primary model, Neuro-Tree-Net, integrates an Artificial Neural Network (ANN) with an Extra Trees (ET) model, combining tree-based learning with neural network-based representation learning to capture complex nonlinear relationships in tourism data. Experimental results demonstrate that the hybrid model outperforms traditional single-model approaches in prediction accuracy. Users can input constraints such as maximum cost, available travel time, and minimum rating preferences to receive personalized destination recommendations. The backend efficiently manages dataset processing, model training, and prediction using libraries including pandas, scikit-learn, Keras, and Matplotlib, while results are presented through a dynamic web interface built with Django. Beyond accurate recommendations, the system supports comprehensive analysis of tourism patterns, enabling travellers to make more informed decisions.