What question did this study set out to answer?

The study aims to evaluate school site suitability using AHP and various ML models in Batna, Algeria.

February 9, 2026Glasnik srpskog geografskog drustva1 citationsOpen Access

Analytic hierarchy process vs. machine learning: Mapping school suitability in Batna city, Algeria with spatial analysis

HTHafida TebbiUniversity of Sciences and Technology Houari Boumediene HDHadda DridiUniversity of Batna 1 ABAbdelmadjid BouderUniversity of Sciences and Technology Houari Boumediene

Key Points

The study aims to evaluate school site suitability using AHP and various ML models in Batna, Algeria.
Utilized Analytic Hierarchy Process for site selection analysis
Implemented three machine learning models: Random Forest, Gradient Tree Boosting, and CART
Conducted spatial analyses using Geographic Information Systems techniques including KNN, Moran's I, and IDW
Identified deficiencies in school distribution across residential sectors
Found that 44.6% of schools are in 52% of residential areas
AHP highlighted population density as a key factor (51.99%)
Random Forest achieved the highest model performance (AUC = 0.75) focusing on environmental factors
GTB and CART showed lower performance (AUC = 0.65 and AUC = 0.61 respectively)
Combined suitability maps offered insights for addressing educational inequalities in Batna.

Abstract

The present study considers the suitability of sites for elementary schools in Batna, Algeria. The Analytic Hierarchy Process (AHP) method is compared to the three machine learning (ML) models, namely Random Forest (RF), Gradient Tree Boosting (GTB) and Classification and Regression Trees (CART). Spatial analysis techniques based on Geographic Information System (GIS) such as K-Nearest Neighbors (KNN), Moran?s I, inverse distance weighting (IDW) and Buffer Analysis were conducted prior to comparison to detect principal deficiencies and showed that 44.6 % of the 83 schools are located in 52% of the residential sectors. Peripheral gaps were also highlighted Nearest Neighbors Ratio (NNR = 0.69, Moran?I = 0.32, p <0.001). AHP prioritized population density: 51.99%, while RF with an Area Under the Curve (AUC = 0.75) emphasized environmental limitations (slope, river spacing), differing from GTB (AUC = 0.65) and CART (AUC = 0.61). The combined suitability maps showed complementarity and guide planners to effectively address Batna's educational inequalities.

اسأل الذكاء الاصطناعي

Bookmark

View Full Paper