CLASSIFICATION AND FORECASTING OF WATER STRESS IN TOMATO PLANTS USING BIORISTOR DATA

Abstract

Drought significantly impacts agricultural productivity, with water stress being the primary concerns affecting the yield. This study leverages real-time in vivo data collected using the Bioristor sensor to classify and predict water stress levels in tomato plants. Water stress affects photosynthesis, transpiration, and nutrient uptake, leading to reduced crop yield and food insecurity. We employ Decision Trees and Random Forest algorithms for classification, distinguishing four stress statuses: Healthy, Stress, Recovery, and Uncertain. Additionally, LSTM-based neural networks are used to forecast future water stress conditions, aiding in automated irrigation management. The proposed system enhances accuracy, efficiency, and water conservation, making it suitable for smart agriculture applications. The Bioristor sensor, an advanced in vivo monitoring tool, provides bio-electrical data that is used to optimize the irrigation strategies. The collected data includes electrical resistance, sap composition, and environmental factors, helping in accurate classification and prediction. Our study compares multiple algorithms, demonstrating that CNN-LSTM hybrid models outperform traditional LSTM in forecasting water stress.