Asian Science Citation Index is committed to provide an authoritative, trusted and significant information by the coverage of the most important and influential journals to meet the needs of the global scientific community.  
ASCI Database
308-Lasani Town,
Sargodha Road,
Faisalabad, Pakistan
Fax: +92-41-8815544
Contact Via Web
Suggest a Journal
Asian Journal of Scientific Research
Year: 2020  |  Volume: 13  |  Issue: 2  |  Page No.: 132 - 139

Incremental Contribution Method to Determine the Size of Neural Network Optimized by Genetic Algorithm

Budi Warsito, Hasbi Yasin and Rukun Santoso    

Abstract: Background and Objective: Determining the optimal architecture and optimization method in neural network for time series modeling have been interesting open problem in recent years. Several studies have been developed to solve this problem, but only to a limited extent. Determination of optimal size with a specific method is only done in the search for weights with standard methods, as well as the use of heuristic optimization without including how to determine the optimal architecture. This paper focuses on the determining optimal architecture in neural network for time series model optimizing by heuristic optimization. Materials and Methods: In this method, the network built first with the big size and then the cells with low contribution, expressed by the R2inc, will be removed from network. To get an approach of global optimum, genetic algorithm was used as optimization method to obtain the optimal weights. This model was applied to the rainfall data. Starting with eight hidden units, NN (1,2,18,5) was chosen as the optimal size of the network, i.e., a network consists of lags 1, 2 and 18 as input and has 5 hidden units. Removing 3 units in the hidden layer is not too large reducing R2. Results: The proposed procedure successfully reduced the size of the network so that the constructed model was simpler. Conclusion: The use of incremental contribution method can effectively reduce network size on neural network optimized by genetic algorithm and is no longer just based on trial and error techniques.

Cited References   |    Fulltext    |   Related Articles   |   Back
  Related Articles

Copyright   |   Desclaimer   |    Privacy Policy   |   Browsers   |   Accessibility