Analyzing Agricultural Land Price Prediction Using Linear Regression and XGBoost Machine Learning Algorithms: A Case Study of Çanakkale

Agricultural lands are known not only as agricultural production areas but also as areas with high income expectations as an investment tool. In Turkey, recent fluctuations in economic indicators such as the euro, dollar, and gold, along with increasing investment demand, have caused agricultural land prices to rise uncontrollably. Controlling land price increases is important for preventing the misuse of agricultural lands. The sustainable management of agricultural lands and price stability are closely related to the United Nations Sustainable Development Goals (SDGs), particularly SDG 11 and 15, “Sustainable Cities and Communities” and “Life on Land.” In this context, accurately predicting prices is important for minimizing price fluctuations in agricultural lands for investors and landowners and supporting sustainable development. In general, the Multiple Linear Regression (MLR) model is considered one of the effective traditional methods for predicting real estate prices. However, depending on the data, more reliable results can be obtained than with powerful deep learning models such as the Extreme Gradient Boosting (XGBoost) algorithm, which exhibits superior prediction performance. This study aims to compare the MLR and XGBoost algorithms to predict agricultural land prices in villages located in the central district of Çanakkale and to examine daily fluctuations in economic indicators such as the dollar, gold, and euro. The results showed that XGBoost (R2 = 0.66) has an advantage in terms of coefficient of determination values compared to MLR (R2 = 0.01). Accurate price prediction for agricultural lands will help control fluctuations in land prices. Additionally, it will support farmers and investors in making informed decisions for a sustainable agricultural economy.