A STATISTICAL EVALUATION OF IMAGE CLASSIFICATION METHODS AND TECHNIQUES FOR ENHANCING THE QUALITY OF LANDCOVER/LANDUSE MAPPING IN AWKA NORTH L.G.A

Accurate classification of landcover/landuse (LULC) is critical for sustainable land management and environmental planning. This study statistically evaluates the effectiveness of various image classification methods -including Support Vector Machine (SVM), Neural Network (NN), Maximum Likelihood (ML), and Random Trees (RT) -in enhancing the accuracy of LULC mapping in Awka North Local Government Area (LGA), Anambra State, Nigeria. Using Landsat 8 OLI imagery, classification was performed after extracting Regions of Interest (ROIs) and conducting spectral separability analysis. Accuracy assessment metrics such as error matrix, kappa coefficient, correlation coefficient, and z-tests were used to evaluate classifier performance. Results showed that SVM (80.83% accuracy) and NN (75.09% accuracy) outperformed ML and RT. This study recommends the use of advanced machine learning techniques for future LULC mapping due to their superior classification performance in heterogeneous landscapes.