BIO-INSPIRED HYBRID GENETIC–NEURAL ARCHITECTURE FOR SMART AGRICULTURE: FERTILIZER RECOMMENDATION AND YIELD PREDICTION

Abstract

Precision farming needs effective resource management to achieve sustainable crop yield. Excessive fertilizer usage leads to deterioration of the soil and the environment. Insufficient fertilizer usage decreases crop productivity and returns on investment. This paper introduces a bio-inspiration-based hybrid Genetic Algorithm and Neural Network (GA-NN) structure for fertilizer recommendation and yield prediction simultaneously. The GA tunes fertilizer quantities, chooses features, and tunes neural network parameters with a multi-objective fitness function. This function reduces yield prediction error and punishes high fertilizer application. Trained parameters are then utilized to build an accurate yield estimation neural network. Agricultural data experiments reveal that the novel hybrid architecture results in fertilizer usage reduction of up to 15% and yield prediction precision improvement of 10% compared to standard machine learning algorithms. The results reflect the ability of bio-inspired hybrid structures to facilitate data-driven and sustainable smart farming policies.