Paramenter Variation Analysis of Learning Vector Quantization Artificial Neural Network For Odor Data Pattern Recognition

Abstract

Abstrak

 

Pengenalan pola data odor dapat dilakukan dengan menggunakan metode Learning Vector Quantization (LVQ) Artificial Neural Network (ANN). ANN dibuat menyerupai sistem syaraf manusia, disebut juga Jaringan Syaraf Tiruan (JST). Dengan beberapa parameter pada LVQ, dapat diketahui karakteristik LVQ sehingga dapat memperkecil error dan epoch, serta memperbesar Recognition Rate. Hasil percobaan menunjukkan hubungan antara parameter alpha, konstanta laju pembelajaran, jumlah epoch, dan inisialisasi vector pewakil terhadap error dan Recognition Rate yang diperoleh.

 

Kata Kunci: ANN, Learning Vector Quantization, epoch, error, JST, Recognition Rate.

Abstract

 

Pattern recognition of odor data can be done by using Learning Vector Quantization Artificial Neural Network (ANN). ANN is made to resemble the human neural system. By varying some parameters on Learning Vector Quantization, Learning Vector Quantization characteristics can be identified to minimize errors and epoch and to enlarge Recognition Rate. The experimental results showed the relationship between the alpha parameter, coefficient alpha, the number of epoch, and vector initialization against error and the Recognition Rate obtained.

 

Keywords: ANN, Learning Vector Quantization, epoch, error, JST, Recognition Rate.