Novel Fine-tuned Bidirectional GRU and FastText Embeddings for Location-based Sentiment Analysis and Predictions
The need for this paper stems from the challenge of efficiently analyzing large volumes of customer reviews in the hotel industry, which is growing and complex due to the widespread use of digital platforms. With consumers increasingly sharing feedback across various social media applications, manual processing becomes impractical, necessitating machine learning algorithms for accurate sentiment analysis and prediction.
This paper addresses the problem by applying machine learning algorithms, specifically fine-tuned bidirectional GRU with FastText embedding, to perform location-based sentiment analysis and prediction of customer reviews in the hotel industry, providing an efficient solution to process large-scale unstructured data and extract valuable insights for improving customer experience.
This study employs machine learning techniques, including Random Forest (RF), Support Vector Machine (SVM), Bidirectional Long-Short Term Memory (BiLSTM), and Bidirectional Gated Recurrent Unit (GRU), to perform sentiment analysis on customer reviews in the hotel industry. The paper focuses on using a fine-tuned bidirectional GRU model with FastText embedding for location-based sentiment analysis and prediction. The research sample consists of a large dataset of customer reviews, which are processed and analyzed to predict sentiment and evaluate model performance, achieving an overall score of 84.31%.
This paper contributes to the body of knowledge by demonstrating the effective application of advanced machine learning algorithms, particularly a fine-tuned bidirectional GRU with FastText embedding, for sentiment analysis in the hotel industry. It offers a unique approach to location-based sentiment analysis, allowing for a deeper understanding of regional variations in customer perceptions. The study also provides insights into the comparative effectiveness of different machine learning models for sentiment classification and prediction by comparing multiple algorithms. This work enhances the existing methodologies for processing large-scale, unstructured data and highlights the potential of sentiment analysis to drive data-informed strategies in customer-centric industries.
The study found that the fine-tuned bidirectional GRU model with FastText embedding achieved an accuracy of 84.31%, outperforming other algorithms in sentiment classification. It highlighted the effectiveness of location-based sentiment analysis for understanding regional variations in customer perceptions. The approach also demonstrated strong predictive capabilities, aiding data-driven decision-making in the hospitality sector.
Practitioners in customer-focused sectors, particularly hospitality, should consider using advanced models like the fine-tuned bidirectional GRU with FastText embedding for precise sentiment analysis. This approach helps capture regional customer preferences, allowing for tailored services. Integrating sentiment insights with real-time data can further enhance responsiveness to customer needs, supporting data-driven improvements in customer satisfaction.
Researchers should investigate more advanced models and varied datasets to enhance sentiment analysis accuracy and applicability. Extending sentiment analysis to fields like healthcare and finance could offer broader insights into consumer behavior. Additionally, integrating sentiment analysis with real-time data sources, such as social media, may yield more dynamic predictive models and reveal key regional trends.
The findings of this paper have significant implications for industries, particularly in the hospitality sector, as they provide an effective method for analysing customer sentiment at a granular level. The ability to assess customer opinions through sentiment analysis can help businesses improve customer service, tailor their offerings to meet customer needs, and enhance the overall customer experience. Additionally, location-based sentiment analysis can enable businesses to understand regional differences, allowing for more personalized marketing strategies and better resource allocation.
Future research could focus on improving the accuracy and efficiency of sentiment analysis models by incorporating more advanced deep learning techniques and larger, more diverse datasets. Exploring the application of sentiment analysis in other sectors, such as healthcare, education, and finance, could provide broader insights into customer perception. Additionally, integrating sentiment analysis with other data sources, such as social media and customer feedback, may lead to more comprehensive and real-time prediction models. Further directions include multilingual sentiment analysis, the use of transformer-based models, and deployment in resource-constrained environments.
Back