This study employs a 4-input and 1-output feedforward neural network with adalines used to implement learning via error back-propagation (EBP) using least mean square rule. The neural network is used to predict the condition of both cooked and uncooked food as well as fresh vegetables by determining food age (in days). Neurosolutions training software is used to simulate the neural network. Training data is obtained from a constructed metal oxide semiconductor (MOS) ammonia circuit. Results show that a 95% overall accuracy of neural network results is obtained. This demonstrates the capability of neural networks in accurate classification of sample data points. Food samples used to obtain inference database include rice, beans, fresh vegetables, yam and potatoes.
Keywords/Index Terms: neural network, supervised learning, back propagation, e-nose, artificial intelligence