Selection of Most Appropriate Backpropagation Training Algorithm in Data Pattern Recognition
This work helps practitioners select effective training algorithms for neural networks in data pattern recognition, but it is incremental as it compares existing methods.
The researchers tested 12 backpropagation training algorithms for pattern recognition in data, finding that the trainlm algorithm achieved the highest suitability at 87.5% with a significance level of 0.000.
There are several training algorithms for backpropagation method in neural network. Not all of these algorithms have the same accuracy level demonstrated through the percentage level of suitability in recognizing patterns in the data. In this research tested 12 training algorithms specifically in recognize data patterns of test validity. The basic network parameters used are the maximum allowable epoch = 1000, target error = 10-3, and learning rate = 0.05. Of the twelve training algorithms each performed 20 times looping. The test results obtained that the percentage rate of the great match is trainlm algorithm with alpha 5% have adequate levels of suitability of 87.5% at the level of significance of 0.000. This means the most appropriate training algorithm in recognizing the the data pattern of test validity is the trainlm algorithm.