Personality-based Restaurant Recommendation
Date Issued
August 1, 2024
Author(s)
Advisor
Abstract
Recommendation systems became ubiquitous in information systems, effectively mitigating decision-makers’ information overloading problem. However, their application in complex tasks, such as restaurant recommendations, faces challenges. This research aimed to address such limitations by investigating the impact of personality (user/venue) and preference on restaurant recommendations. The study employs deep learning and machine learning approaches to extract personality traits and preferences from electronic word-of-mouth (e-WOM). Various data imbalance treatment techniques are explored to optimize the performance of the models. Furthermore, a customized Named-Entity Recognizer (NER) is developed to extract users’ food preferences from text, improving traditional approaches that expect users to manually input these in the system. The proposed approach combines topic modelling, text classification and named-entity recognition, to extract users’ personality, preferences, and opinions from e-WOM, and train an eXtreme Gradient Boosting (XGBoost regressor) model, to predict user satisfaction with restaurants. The XGBoost predictions are used as a proxy to recommendation. The approach is compared against traditional techniques (e.g. collaborative filtering) using different recommender systems performance metrics. The findings show the advantage of the proposed approach over traditional techniques. The results also contribute to a better understanding of the role of personality(user/venue) in restaurant recommendation. The approach has implications for both researchers and practitioners in the fields of recommender systems.
File(s)![Thumbnail Image]()
Name
phd_evros_christodoulou_2024.pdf
Size
1.78 MB
Format
Adobe PDF
Checksum (MD5)
8148d33de71cb98368dfc471b1f697f6