Virtual analysis of machine learning models for diseases prediction in muskmelon

Abstract

Muskmelon, a crop prized for its economic potential, has a relatively brief growth cycle. Disease susceptibility during this period can have a profound impact on yields, posing challenges for farmers. Environmental conditions are pivotal in disease occurrence. Unfavorable conditions reduce the likelihood of pathogens infecting vulnerable host plants as temperature and humidity influence pathogen behavior, including toxin synthesis, virulence protein production, and reproduction. Pathogens can lie dormant in the soil until suitable conditions activate them. When the right environment and host plants align, these dormant pathogens can cause outbreaks. Disease prediction becomes possible by analyzing environmental variables. Real-time data collected via strategically placed sensors focused on viral, fungal, and bacterial infections. Results indicated that the extreme gradient boosting (XGBoost) algorithm, with a maximum tree depth of 4 and 30 trees per iteration, achieved remarkable performance, yielding an accuracy of 97%. For comparison, the XGBoost model outperformed an 8-layer Backpropagation network with 7 nodes per layer, which achieved 95% accuracy. These findings underscore XGBoost's efficacy in forecasting and mitigating muskmelon plant diseases, offering the potential for improved crop yields and agricultural sustainability.