Research on fusion algorithm with BP neural network and rough set
Considering that the neural network increases rapidly in complexity and is greatly extended in training time due to the gradually increasing input vectors, a fusion algorithm modified by rough set is proposed to preprocess the inputs before neural network. Firstly, the input sample space is reduced...
Saved in:
Published in | 2011 6th IEEE Conference on Industrial Electronics and Applications pp. 118 - 123 |
---|---|
Main Authors | , , , , |
Format | Conference Proceeding |
Language | English |
Published |
IEEE
01.06.2011
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | Considering that the neural network increases rapidly in complexity and is greatly extended in training time due to the gradually increasing input vectors, a fusion algorithm modified by rough set is proposed to preprocess the inputs before neural network. Firstly, the input sample space is reduced to obtain the new decision table according to attribute significance in rough set. Then the reduced decision table is applied to train the BP neural network until it converges. Finally, the algorithm is used for classification in car mileage level. The classification accuracy of the testing sample is increased by 10% than the traditional BP neural network fusion algorithm. The result shows the modified fusion algorithm is feasible. |
---|---|
ISBN: | 9781424487547 1424487544 |
ISSN: | 2156-2318 2158-2297 |
DOI: | 10.1109/ICIEA.2011.5975561 |