DEVELOPMENT OF AN ARTIFICIAL INTELLIGENCE BASED EVENT RECOMMENDATION SYSTEM FOR EVENT PLATFORM

Abstract

On event platforms, users often try to find events through search functions, and this process is insufficient for matching user with right event. In this study, a dynamically learning and continuously updated, machine learning-based multi-layered activity recommendation system has been developed, enabling users to discover activities on platform faster, more accurately, and in a personalized way. Within the scope of study, different recommendation models have been developed, utilizing Collaborative Filtering (CF), Content-Based Filtering (CBF), and deep learning-based filtering. In CF, Singular Value Decomposition (SVD), k-Nearest Neighbors (KNN), and Non-Negative Matrix Factorization (NMF) have been used to learn user-product interaction patterns. In CBF approach, Term Frequency–Inverse Document Frequency (TF-IDF) and Cosine Similarity have been used to identify similarities based on event characteristics. In hybrid models, Weighted, Cascade, and Switching approaches have been used to combine CF and CBF models. In deep learning, Neural Collaborative Filtering (NCF), Autoencoder, and Long Short-Term Memory (LSTM) have been used to model nonlinear and temporal relationships. LSTM model demonstrated most successful performance, achieving the highest results in all evaluation metrics. With developed system, the event discovery process has become faster, more efficient, and more satisfying for users. As a result, user satisfaction has been maximized.