Miglioramento della pubblicità mirata fuori casa a Manhattan Basato sui dati dei social media e sul modello BERTopic

The research assessment and study essentially aims to explore the relationship between mostly traditional forms of advertising, and between largely contemporary forms of geo-tagged data, ad-personalization and targeted advertising, which are mostly utilized within social media marketing spaces. In order to achieve that 78,347 geo-tagged Tweets posted between the dates 1st of June, 2021 and 30th of September, 2021 over 10 popular neighborhoods or areas in lower Manhattan are collected. Generally, there are specific models that are being in topic modelling, one of them is Latent Dirichlet Allocation (LDA) and the other one is Non-negative Matrix Factorization (NMF) which more of a statistical method that learns by decomposing the text corpus. In this research, both methods showed poor results, and a pre-trained model called BERTopic is used. In terms of interpretable results, BERTopic showed much better results. From the ten neighborhoods and areas, a total of 12 different topics were established, the vast majority of which were developed by way of a prevalent and high-volume number of keywords and search terms that showed specific interest in the according trends and patterns. These topics are “Food & Dining,” “Parks & Recreation,” “Tourism,” “Fashion,” “Beauty & Care,” “Bespoke and Formal Attire,” “Education & Student,” “Architecture & Design,” “Science & Technology,” “Reading,” “Lifestyles & Relationships,” “Organic Produce.” Altogether, the data and findings clearly shows that the vast majority of each of the ten neighborhoods or areas possesses unique topic groupings, which can be translatable into marketable segments, thus targeted out-of-home advertising.