SMART URBAN AGRICULTURE: AUTOMATED NUTRIENT MANAGEMENT AND PEST FORECASTING

Abstract

Urban agriculture is increasingly being adopted as a sustainable method to improve food availability in rapidly urbanizing environments. Despite its benefits, managing nutrients effectively and responding promptly to pest threats continue to be significant hurdles in maximizing crop output within space-constrained urban settings. This research proposes a Smart Urban Farming System that leverages machine learning to automate fertilizer selection based on real-time environmental and soil conditions. The system processes multiple input variables, including temperature, humidity, moisture content, macronutrient levels (N, P, K), crop type, and soil classification, to predict the optimal fertilizer required for plant health. Two machine learning models are employed—Logistic Regression (as a baseline) and Random Forest (as the proposed advanced solution). Performance comparison reveals that the Random Forest model achieves an impressive accuracy exceeding 91%, far outperforming the baseline Logistic Regression model, which scored only 15%. Evaluation through metrics such as precision, recall, and F1-score reinforces the reliability and robustness of the proposed model. By automating nutrient recommendations, the system reduces manual intervention, increases efficiency, and supports environmentally conscious farming practices. Furthermore, the model is designed with future extensibility in mind, allowing seamless integration with IoT-based sensing systems for real-time automation and potential expansion into pest risk prediction and full-scale urban farm control. This work contributes to the development of intelligent, scalable solutions for sustainable urban food systems