Modelling Maintenance Cost Prediction for Public Office Buildings in North West Nigeria

Reliable prediction of maintenance costs for public office buildings is critical for informed long-term budgeting, asset management planning, and government financial management in Nigeria. This study develops a maintenance cost prediction model for public office buildings in the North West geopolitical zone, using Kano, Kaduna, and Sokoto States as the study area. An archival quantitative design was employed, with maintenance cost records collected from 55 public office buildings maintained between 2018 and 2024 supplemented by survey data from 142 facility managers and quantity surveyors. Multiple regression analysis and artificial neural network modelling were applied to identify cost drivers and develop a predictive model. Findings show that building age, floor area, construction quality rating, and air conditioning system complexity are the strongest predictors of annual maintenance costs, explaining 67% of cost variance in the regression model. The neural network model achieves a predictive accuracy of 88.3% on the validation sample, outperforming the regression model on non-linear cost patterns. Buildings over 20 years old in the region incur maintenance costs averaging 3.8% of replacement value per annum, significantly above the 1.5% typically budgeted. This study fills a specific methodological gap in Nigerian maintenance cost modelling by comparing regression and machine learning approaches in a public building context. Recommendations address budget reform, building condition survey protocols, and model implementation by state ministries. Keywords: maintenance cost prediction, public office buildings, North West Nigeria, machine learning, cost modelling