Time series data modelling for classification of drought in tomato plants

Plants exhibit a regulated physiological activity rendering to various environmental changes. The physiological activity of a plant is recorded in the form of electrical potentials, playing a significant role in understanding plant processes. The electrical signal studies present various innovative methods to exactly detect and classify the state of the plant. In this study, two different databases of tomato plants are used for modelling the development and testing, where Database 1 was created by acquiring signals from tomato plants with help of BIOPAC®MP150 in healthy and drought conditions, before and after visual symptoms were portrayed. Data were visualized through polar plots and signals from different plants and were grouped into non-repetitive combination pairs (fits) to extract shape coefficients. Subsequently, shape coefficients of the signals were extracted from each fit using curve-fitting polynomial models in three different degrees of order. Binary classification was performed for each fit to classify healthy and drought signals, and Ensemble Boosted Tree machine learning classifier was used for classification. Reported results achieved the highest fit-wise classification rate of 97.9 and 86.9% in two databases. The testimonies portrayed a high classification rate before the appearance of visual symptoms in tomato plants. Performing such studies to predict drought effects in pre-symptom stages for drought for various tomato plants provides a reliable innovation to develop preventive measures to protect plants from environmental effects and help crop protection practices.