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

Abstract

Abstract - The main goal of this study is to describe, classify, and predict water stress in tomato plants by using real-time data from a new monitor called the bioristor and different AI models. At first, classification models like Decision Trees and Random Forest were used to tell the difference between tomato plants that were in different levels of stress. We used Recurrent Neural Networks (RNNs), especially Long Short-Term Memory (LSTM) networks, to guess how much water tomatoes would need in the future, taking into account both one- or more-state situations. The results showed that the bioristor sensor and AI models work well in real-world smart watering systems by showing high accuracy, precision, memory, and F-measure. This study builds on the approach used in the base paper by adding more methods, like Convolutional Neural Networks (CNN) and a Voting Classifier, to the analysis. It was able to achieve an impressive 97% accuracy. The study also shows that success can be improved by using ensemble methods, which combine estimates from different models. A frontend built on the Flask framework with user authentication is also suggested to make testing easier for users. Overall, this study shows how modern devices and machine learning can be used to improve farming output and make watering more efficient. Index Terms: AI modeling and forecasting, bioristor, precision agriculture, recurrent neural network, tomato plants, tree-based classifiers, smart irrigation, water stress.