Development of a Sentiment Analysis-Based Proposal Scheme for Tourist Destinations Using a Novel Random Forest and Support Vector Regression (RF-SVR)

Abstract

Introduction: Tourism is a vital sector that relies on customer satisfaction and feedback; therefore examining tourists' opinions of various areas may benefit local governments and businesses. Decision-making systems in numerous sectors need sentiment analysis approaches.
Methods: This research describes a novel technique for sentiment analysis for tourist destinations that combines two Machine Learning (ML) approaches: Stochastic Random Forest-Dynamic Support Vector Regression (SRF-DSVR). This scheme attempts to deliver personalized location suggestions by analyzing traveler evaluations, collecting sightseeing reviews, ratings, and weather data for varied classifications, and improving sentiment analysis through data collection and augmentation. We use tokenization and stop word removal to clean and prepare the text data. The Term Frequency-Inverse Document Frequency (TF-IDF) approach converts text input into numerical vectors, allowing useful characteristics for ML algorithms to be extracted. The SRF-DSVR system successfully determines tourist preference uncertainty and variability by offering robust sentiment analysis and reacting to changing sentiment patterns, allowing the system to deliver current suggestions that correspond with current tourist preferences.
Results: The findings reveal that the SRF-DSVR combination is superior to standard sentiment analysis algorithms in f1-score of 94,2 %, accuracy of 95, 8 %, precision of 94, 9 %, and recall of 95, 3 %. We also performed a detailed assessment utilizing confusion matrices, which revealed the model's efficiency in sentiment classification tasks.
Conclusion: The research demonstrates the potential of the SRF-DSVR technique for sentiment analysis and suggestion schemes in tourist sites, therefore improving visitor experiences and allowing local governments and companies to make data-driven decisions.