Fuzzy Logic Recommender Model for Housing
Emanuel Guillermo Muñoz Muñoz, Jaime Meza, Sebastián Ventura · IEEE Access · 2025
Recommending suitable housing faces significant challenges due to the continuous increase in demand and the need to meet habitability standards. This document presents an innovative approach to address these challenges through a housing recommendation method based on distances to key spatial points and latent characteristics of the properties. The proposed method uses objective distances from the properties to points of interest, such as educational centres, medical centres, pharmacies, shops, entertainment, 911 security cameras, and public transport stations. These distances are calculated based on the area where the property is located, providing an accurate assessment of the environment. Additionally, housing features are grouped into three correlated latent factors: Size and Value, Environment and Comfort, and Age and Safety. The recommendation system relies on fuzzy control to manage user preferences and select appropriate input data to test the model. Initially, a content-based filtering approach is used, as housing ratings are unavailable. The model predicts a percentage of membership in each cluster, which allows for handling uncertainty by offering properties from different groups proportionally. Euclidean distance measures the similarity between user preferences and housing characteristics. Then, search time is optimized through metaheuristic methods, with the bat algorithm offering the best performance in terms of time. This algorithm selects the properties displayed to the user based on natural features extracted from real estate platforms through web scraping techniques. The system is built with a Model-View-Controller architecture using Python, Flask, and SQLite. Personal customer data is also recorded to create clusters and calculate distances with new customers, allowing properties with high ratings to be recommended. This approach combines collaborative and content-based filtering, creating a hybrid system that improves recommendation accuracy and relevance. This analysis shows that the new recommendation method is an effective and accessible solution for selecting suitable housing.