The Use Of Machine Learning In Estimating The Housing Deficit: A Systematic Literature Review

Abstract

ABSTRACT

Objective – Identify studies that apply techniques related to Artificial Intelligence — including machine learning, deep learning, and data mining — to the analysis of the housing deficit or similar tasks based on census data.

Methodology – The studies are identified through bibliometric and systematic reviews of the scientific literature.

Originality/Relevance – With recent advances in computational methods, especially machine learning algorithms, there is interest in evaluating their potential to deal with the complexity of the housing deficit phenomenon and support the formulation of more effective urban and housing policies.

Results – Using the Scopus database, 1,528 documents published between 1985 and 2024 were identified that combine computational methods and census data, of which 18, published between 2015 and 2024, directly address issues related to the housing deficit.

Theoretical/Methodological Contributions – The results indicate the predominance of tree-based algorithms, particularly the Random Forest method.

Social and Environmental Contributions – the diversity of application areas and data types used limits the identification of an ideal model, highlighting the need for further exploration of the approaches and contexts analyzed.