Context aware point of interest recommendation over spatio temporal features in location based social network

Abstract

The emergence of Location-based Social Networks (LBSNs) has transformed how people newlineengage with their environment by allowing easy sharing of locations, experiences, newlineand preferences. They have accelerated the advancement of Point-of-Interest (POI) newlinerecommendation systems, providing users with tailored recommendations for locations newlineand things to do. Yet, creating precise and relevant POI suggestions is a difficult task newlinedue to the need to address issues like sparse data, privacy concerns, and the requirement newlinefor contextual significance. newlineTo alleviate these issues, this thesis suggests a number of innovative approaches. newlineUser-POI embeddings and Laplacian noise are used in a differentially private approach newlineto provide safe social network discovery while protecting user privacy. To capture the newlinechanging nature of user preferences, spatiotemporal proximity and dynamic preference newlinemining are used to improve contextual modeling. Additionally, multi-agent techniques newlineemphasizing fairness, consensus, and concession are incorporated into tactics for group newlinePOI suggestions. The results demonstrate that the suggested approach is highly effective newlineat reducing data sparsity using sophisticated modeling and embedding strategies, newlineensuring that the recommendations are not only precise but also varied and coincidental. newlineBy lowering the number of reiterating recommendations and exposing users to new newlineplaces that suit their changing preferences, this variety improves user happiness. Moreover, newlinethe use of spatial and temporal dynamics guarantees that the recommendations newlinemaintain contextual relevance, providing users with tailored ideas that adjust to their newlinecurrent situation. newlineThis study has real-world implications in e-commerce, smart cities, tourism, and newlineiii newlinehealthcare, where location-based suggestions that are both customized and privacyconscious newlineare essential. In order to further improve POI recommendation systems, newlinefuture research will examine adaptive learning processes, cross-domain recommendations, newlineand sophisticated models using federated