Developing a web application for co-living

Abstract

The article examines the problem of finding neighbors for joint rental housing in conditions of high population migration to large cities and the lack of specialized selection tools. The relevance of the study is due to the growing demand for shared rentals and the low efficiency of existing search methods through social networks, where ads are presented in an unstructured form and require significant time for analysis. The scientific novelty of the work lies in the development of a specialized web application that combines mechanisms for structured ad storage, intelligent filtering and candidate ranking based on an assessment of the similarity of user profiles. An approach to building a recommendation system is proposed, including preliminary filtering of incompatible users and the subsequent calculation of a complex similarity function using categorical, multiple and numerical features. Special attention is paid to the process of aggregating data from social networks, the use of LLM models for semantic analysis of text descriptions of posts and automatic extraction of user characteristics without the need to form a training sample. The assessment of the quality of categorical feature extraction is considered through the prism of a multiclass classification.