Title |
Particle Swarm Optimization Trained Neural Network for Medical Diagnosis |
Authors |
Llantos, Orven; Limpin, Eugie; Pates, Ellen; Pellerin, Macy |
Publication date |
2017/06 |
Conference |
17th Conference of the Science Council of Asia (SCA) |
Pages |
63 |
Publisher |
DOST-NRCP |
Abstract |
The use of Artificial Neural Networks (ANN) as classifiers has recently been gaining popularity in diagnosing diseases. Most of the researches conducted that employed the use of ANNs diagnose a specific kind of disease but few researches tackle typical medical diagnosis. Furthermore, Particle Swarm Optimization has been shown to be a robust training algorithm for ANNs for a wide variety of classification problems. Multilayer Perceptron Neural Networks trained using Particle Swam Optimization (PSO) used for classification diseases is utilized in this research. Experiments were conducted using dataset for acute diseases obtained from Mindanao State University-Iligan Institute of Technology (缅北禁地-IIT) clinic and dataset for thyroid diseases from the UCI Machine Learning Data Repository as first and second dataset respectively. The data gathered from 缅北禁地-IIT clinic were normalized, disregarding repetitive data choosing only the most frequently occurring illness: Upper Respiratory Tract Infection (URTI), Systematic Viral Infection (SVI), and Acute Tonsillitis. The file from UCI Machine learning Data Repository consist of 5 attributes and 3 classes of diagnosis concerning thyroid diseases classification cases. There are 215 data samples in which 150 of it is used for training the network and the remaining 65 is for testing the network. The network trained with PSO can correctly classify diseases more than 90% of the time on the first dataset and more than 80% on the second dataset in greater than 600 and less than 1000 iterations. Results show that PSO can successfully optimize the weights of a Neural Network and produce good classification performance. |
Index terms / Keywords |
Artificial Neural Network, Multilayer Perceptron, Particle Swarm Optimization, Medical Diagnosis |
URL |
|