An Analysis of Kohonen Algorithm Addition to the Backpropagation Method in Processing of Recognizing Temperature Data
A Hasibuan1, M Zarlis2, H Mawengkang3, P H Putra4
1A Hasibuan, Graduate, School of Computer Science, Universitas Sumatera Utara, Malaysia.
2M Zarlis, Graduate, School of Computer Science, Universitas Sumatera Utara, Malaysia.
3H Mawengkang, Graduate, School of Mathematics, Universitas Sumatera Utara, Malaysia.
4P H Putra, Graduate, School of Mathematics, Universitas Sumatera Utara, Malaysia.
Manuscript received on 09 May 2019 | Revised Manuscript received on 19 May 2019 | Manuscript Published on 23 May 2019 | PP: 1359-1361 | Volume-7 Issue-6S5 April 2019 | Retrieval Number: F12360476S519/2019©BEIESP
Open Access | Editorial and Publishing Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Climate change is a global phenomenon triggered by an increase in the average temperature of the earth air layer as the greenhouse gasincreases in the air layer. This global temperature change has been impacted on climate change. The results of this study: With the addition of the Kohonen algorithm in the back propagation method, it is faster to reach the target error if compared by using random weights. An Decreasing of squared error in the training of back propagation method with a error target of 0.009 with random weight reach 37 iterations, while the addition of Kohonen weight in the back propagation method reaches the error target of 36 iterations. The results test with the addition of kohonen in back propagation method with a error target, 0.01, 0.0095, 0.009, 0.008, 0.007 is no better than using random weights. The testing data achieves better accuracy target error 0.009 reaches 95.73% accuracy.
Keywords: Neural Network, Kohonen, Backpropagation.
Scope of the Article: Neural Information Processing