Forecasting pest risk level in roses greenhouse: Adaptive neuro-fuzzy inference system vs artificial neural networks

Abstract

The purpose of this study is to establish a system for the prediction of the pests’ risk level in a roses greenhouse by applying Artificial Neural Networks (ANNs) and an Adaptive Neuro-Fuzzy Inference System (ANFIS). Pests in roses greenhouses are known to be fatal to plants if not detected at a premature stage. Early detection could avoid huge agronomic and economic losses. Though, it could be a difficult task to achieve. The complexities arising from the interactions between variables influencing the development could be a barrier to fulfill the previously mentioned task. The output of the developed system represents the next day?s risk level of Western flower Thrips (WFT) (Frankliniella occidentalis) in a roses greenhouse. Four explanatory variables, such as internal temperature, internal humidity, today’s pest risk level and human intervention have been considered for this estimation. The main contributions of this study are three fold; providing a daily estimate WFT risk level, reducing the use of pesticides and finally mitigating yield loss. The obtained results were compared to each other and to real data. The performance of the models has been evaluated by 3 statistical indicators. Numerical results showed conspicuous performance of both models, indicating their efficiency for pest monitoring. The novelty associated with the system is the creation of decision support tool for daily risk assessment of WFT. Relying on a small number of variables, this system is a monitoring tool which contributes to help farmers early reveal warning signs. In addition, this is a first attempt to employ ANNs and ANFIS for the prediction of WFT.

Keywords

• Decision making
• Artificial neural networks
• ANFIS
• Risk assessment
• Integrated pest management