A Multi-tiered Solution for Personalized Baggage Item Recommendations using FastText and Association Rule Mining: A Novel Approach

Abstract The use of automated systems is helpful for the social transformation to an advanced society. This research paper presents a baggage item recommendation system that leverages FastText word embeddings and Association Rule Mining (ARM) to deliver personalized suggestions to users. The methodology involves several key steps. First, the baggage data is collected and preprocessed, including loading pre-trained Fast-Text word embeddings for text representation. The Phase-I recommendation system is then developed, using FastText embeddings to calculate similarity scores between user input and existing items. Next, the Phase-II recommendation system is extended to incorporate user search history by recording searches and storing them in a data frame. In Phase-III, ARM is applied to the user search history, generating associationrules that enrich the recommendation process. The final stage, Phase-IV, combines FastText and ARM-based recommendations, providing users with more accurate and relevant suggestions. The research evaluates the recommendation model’s performance using metrics such as coverage, support, confidence, lift, leverage, and conviction. The proposed approach achieves enhanced item recommendations, offering insights intouser behavior and preferences, and opens opportunities for personalized recommendations, marketing strategies, and product optimization.