non-professional personnel, thus it does not apply to non-commercial farmers. This paper implements a cost-effective pest detection method that can be used for microfarmers by using a low-cost electronic nose and machine learning modelling. This method was realized by conducting a controlled experiment and making measurements, we designed and conducted controlled experiment using wheat and an oat aphid as experimental materials, while measurements were made using E-nose and an open gas exchange system to acquire data sets used for model construction. Artificial Neural Network (ANN) was used to develop three models including two classification models and one regression model. These classification models are able to classify the level of pest infestation based on E-nose measurements and the regression model can predict three physiological parameters including photosynthesis, stomatal conductance and transpiration based on E-nose measurements. Both classification models achieve high accuracy around 98% and the best one of them was confirmed through model evaluation and analysis to have no over-fitting or under-fitting problems. The regression model has the performance of overall correlation coefficient 0.79. Model evaluation were analyzed based on accuracy, Mean Squared Error (MSE) and correlation coefficient. These three models enable the method implemented in this study to detect pests efficiently and reliably.