Evaluating the Performance of Machine Learning Algorithms and Maximum Likelihood Classifier for Land-Use and Land-Cover Change Detection in Yola, Nigeria

Muhammad Isma’il, Sule Muhammad Zubairu, Auwal Aliyu, Muhammad Hadi Ahmed, Sadiq Ibrahim, Abubakar Magaji and Idris Moisule Hassan

Keywords: Classification, Land use Land cover (LULC) Change, Machine learning, Support vector Machine, Artificial Neural Network, Maximum likelihood, classifier.


The availability of high-end consumer computing power and free satellite data has caused data science and remote sensing groups to begin aligning in recent years. Openly accessible data from the Landsat series of the United States Geological Survey (USGS) have been used in a number of remote sensing applications. Nevertheless, there is a lack of studies that utilize these data to evaluate the performance of machine learning algorithms and maximum likelihood for LULC classification in the complex Yola-North Landscape. In this study, we compared the classification performance of support vector machines (SVM), Artificial Neural Network (ANN) machine learning algorithms and maximum likelihood parametric algorithm. The study area chosen is a complex mixed-use landscape with four major land use and land cover (LULC) classes built-up, vegetation, water and bare soil. Multi-temporal scenes from Land sat 7 ETM+ and Land sat 8 OLI images covering the period of 2002, 2012 and 2022 were used for the classification. Accuracy was assessed using Overall Accuracy and Kappa Coefficient derived from an error matrix. The Land Change Modeler (LCM) was used to detect changes in the LULC. The results show that the highest overall accuracy was achieved by support vector machines (95.5%, 87.7% and 90%) for year 2022, 2012 and 2002 respectively, closely followed by Artificial Neural Network (89.5%, 87.5% and 85%) for 2022, 2012 and 2002 respectively, and finally, Maximum Likelihood (between 84.5% and 70%). The findings also show that, the built[1]up areas have increased significantly, this increase coincided with a decrease in vegetation area, bare surface and water, and without proper planning it could cause a severe consequence for the environment, society, and overall sustainability. By prioritizing sustainable development, integrating smart growth principles, and fostering collaboration among stakeholders, it is possible to mitigate these negative effects and create more resilient, livable, and equitable urban environments.