Agarwood oil quality classification using cascade-forward neural network

Agarwood, also known as Gaharu in Malaysia, is a fragrant and valuable international commodity harvested from Aquilaria and Gyrinops tree species. The quality of agarwood depends on many factors, such as the quality of its wood resin, smell and origin. Current methods for determining its quality rely on human experts. However, an automated approach would be more suitable for mass production. In this paper, we propose the Cascade Forward Neural Network (CFNN) to perform agarwood oil quality classification. Gas Chromatography-Mass Spectrometer (GC-MS) samples collected by Forest Research Institute Malaysia (FRIM) and University Malaysia Pahang (UMP) were used to train a CFNN to classify the quality of the agarwood. The hidden units and output threshold were varied to determine the optimal model. The results show that the optimal CFNN (with 1 hidden node and 0.5 threshold) managed to obtain 100% classification accuracy on the dataset.