PENGARUH PRINCIPLE COMPONENT ANALYSIS TERHADAP TINGKAT IDENTIFIKASI NEURAL NETWORK PADA SISTEM SENSOR GAS
Abstract: In recently, it has
been developed a gas identification system consists of a semiconductor sensor
array and Neural Network pattern recognition. In this study, it has been
implemented a method of Principle Component Analysis (PCA) as a preprocessing
of the Neural Network algorithm. The sensory array is composed of eight
commercial semiconductor sensors. Three layer-Neural Network was trained with
the back propagation technique within 5000 epochs. PCA could reduce the
eight-dimension into three-dimension without any information losses. The identification error rate was lower with
the ratio of ~10-4 and the training period was shorter with the ratio of ~0.6.
In generally, it can be concluded that the implementation of the PCA method
into the Neural Network can enhance the performances of the neural include the
identification rate and time consumed in the training phase.
Penulis: Muhammad Rivai
Kode Jurnal: jptkomputerdd070043
