Urban Land Property Valuation for Taxation Purpose Using GIS and Deep Learning: The Case of Adama City, Adama, Ethiopia

Abstract

Urban land property valuation is a critical element of efficient and equitable property taxation systems, particularly in rapidly growing cities such as Adama, Ethiopia. This thesis develops an integrated framework combining Geographic Information Systems (GIS) and deep learning techniques to enhance the accuracy and transparency of urban land valuation for taxation purposes. The study utilized diverse datasets, including high-resolution satellite imagery, cadastral maps, road networks, infrastructure layers, and socio-economic attributes, sourced from the USGS, Copernicus Global Land Cover, OpenStreetMap, and local authorities. Methodologically, the study adopted a two-pronged approach. First, a GIS-based Multi-Criteria Evaluation (MCE) was implemented using the Analytic Hierarchy Process (AHP), where parameters such as road proximity, utility access, and service availability were weighted?��?with road proximity receiving the highest weight (30%). Second, a deep learning model based on Convolutional Neural Networks (CNNs) was developed using Python and TensorFlow in Google Colab. The CNN model was trained on over 1,000 parcel samples and achieved a mean squared error (MSE) of 0.021, demonstrating high prediction accuracy for continuous land value estimation. The final land valuation ma"categorized parcels into five value classes (Very Low to Very High). Approximately 24.6% of the parcels fell into the "Very High" value class, predominantly located in the central business districts and near primary roads and service areas, while 29.3% fell into the "Very Low" category, primarily in peripheral or under-serviced zones. The integration of CNN-based predictions and GIS-based spatial modeling significantly improved the granularity and reliability of the land value classification. The study concludes that combining deep learning with GIS enhances the precision, objectivity, and scalability of urban land valuation. This approach enables more equitable taxation, optimized infrastructure investment, and efficient land administration. The research recommends adopting AI-powered valuation