International Journal of Statistics and Applied Mathematics
  • Printed Journal
  • Indexed Journal
  • Refereed Journal
  • Peer Reviewed Journal

2023, Vol. 8, Special Issue 1

A comparative study of different multi-layer preceptorn learning algorithm with raw and normalized for classification of methane and hydrogen


Author(s): Maibam Sanju Meetei

Abstract:
In this study, the classification of 0.5% methane gas, 1% methane gas, 0.5% hydrogen gas, and 1% hydrogen gas is done using the multi-layer perceptron neural network. The neural network with one hidden layer is sufficient to classify the data in theoretically but it is number of neurons, training period is high and low convergence rate. So a neural network with two hidden layer is consider for the study as this network has higher convergence rate with lesser number of neurons. Out of eleven different learning algorithms, Levenberg-Marquardt learning algorithm shows the best algorithm for this study. Additionally, the impact of training with raw data versus training with normalized data is examined, and it is found that training with normalized data performs better because the neural network architecture has less number of neuron and higher performance i.e. mean square error is smaller. This research clearly demonstrates that MLP may be used to classify methane and hydrogen at various concentration levels. For raw data training, the numbers of neurons are fourteen and twelve neurons in first and second hidden layers respectively with a minimum MSE of 0.00586. For normalized data training the numbers of neurons are ten and eleven neurons in first and second hidden layers respectively with a minimum MSE of 0.002.


Pages: 43-47 | Views: 421 | Downloads: 13

Download Full Article: Click Here
How to cite this article:
Maibam Sanju Meetei. A comparative study of different multi-layer preceptorn learning algorithm with raw and normalized for classification of methane and hydrogen. Int J Stat Appl Math 2023;8(1S):43-47.

Call for book chapter
International Journal of Statistics and Applied Mathematics