An LSTM-Driven Model for Accurate Agricultural Pest Risk Prediction

Abstract

This research aims to build an effective predictive modeling system for agricultural pest risk using machine learning techniques and weather conditions. The study focuses on increasing prediction accuracy by checking deep learning models to help with timely and correct pest management decisions. In this research, we saw at two machine learning ways. Group 1 used a CNN model. Group 2 used with an LSTM model to use 9000 samples research weather parameters like temperature, humidity, rainfall, and seasonal designs. We assessed the model's performance with an self t-test at 95% confidence. It shows that the LSTM model outperformed the CNN, achieving 96% accuracy checked to the CNN's 85%. LSTM also showed better precision, recall, and F1-score, and it was more stable and consistent across the results. The statistical analysis approved that the performance difference between the two models was important (p < 0.05), shows the superior effectiveness of the LSTM algorithm in predicting pest risk. The study concludes that LSTM-based predictive modeling provides more accurate and reliable predictions for agricultural pest risk than CNN. It is especially right for early pest detection and decision support in agriculture. This method can help remove crop losses and improve pest management strategies, particularly in data-driven and real-time agricultural systems.