The Selection of Neural Network Input Parameters Based on Association Rules

DOI

10.2991/cmsa-18.2018.72How to use a DOI?

Keywords

neural network; association rules; parameter selection; dimension reduction

Abstract

Neural network has strong ability of nonlinear approximation and fitting, which was widely used in various prognosis prediction researches. Meanwhile, the selection of neural network input parameters was very important: the number of input layer nodes would increase as the number of input parameters and it required a large number of sample data to train the neural network, which was very easy to cause the dimension disaster. Otherwise, the sample data and the local extremums in the process of convergence would decrease as the number of input parameters, which could simplify the topology of the neural network prediction model greatly. In this paper, we used association rules based on data mining to select input parameters of the neural network prediction model of cervical spinal cord injury and reduced its dimensionality from 25 to 7. Finally, the good performance is proved through the simulation results using MATLAB.

Copyright

© 2018, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Lei Shi
PY  - 2018/04
DA  - 2018/04
TI  - The Selection of Neural Network Input Parameters Based on Association Rules
BT  - Proceedings of the 2018 International Conference on Computer Modeling, Simulation and Algorithm (CMSA 2018)
PB  - Atlantis Press
SP  - 317
EP  - 319
SN  - 1951-6851
UR  - https://doi.org/10.2991/cmsa-18.2018.72
DO  - 10.2991/cmsa-18.2018.72
ID  - Shi2018/04
ER  -