Classification Learning Assisted Biosensor Data Analysis for Preemptive Plant Disease Detection

Abstract

In the agricultural sector, identifying plant diseases is crucial as they hamper the plant's robustness and health, which play a vital role in agricultural productivity. Early detection allows farmers to take proper measurements and save crops from complete failure. Biosensor applications in agricultural production and plant monitoring improve the yield through definitive recommendations and improved practices. Fluorescence-based assays, colorimetric biosensors, and surface plasmon resonance-based biosensors are the most commonly used for plant pathogen detection. Plant disease detection is a prime biosensor application for preventing seasonal and cultural defects in raising crops. The sensor data analysis ensures reliable processing for distinguishable features for identifying the disease, wherein discrete information is handled with error. This article introduces a Preemptive Classification using Discrete Data (PC-DD) technique to resolve this issue. This technique requires partial series through probabilistic data substitution to improve the analysis rate. In the analysis process, the classification is performed using random forest based on two combinations: series, difference, and probability. This probability is based on identical data observed through series probability classification in the previous iteration. The unidentical data is classified under difference that is used for individual classification. This process is progressive until the detection is performed, wherein the alterations are adaptable for different biosensor input data. Therefore, the proposed technique's performance is validated using the metrics detection accuracy, analysis ratio, analysis time, classifications, and difference factor.