Parkinson Disease Classification based on binary coded genetic algorithm and Extreme learning machine
Vasily Sachnev, Hyoung Joong Kim · 2014
In this paper, we propose a Binary Coded Genetic Algorithm combined with Extreme learning machine (BCGA-ELM) for Parkinson Disease classification problem. Proposed method analyses ParkDB data base of 22283 genes' expression information extracted from 22 normal patients and 50 Parkinson Disease patients. Proposed method can sufficiently recognize PD patients among normal persons using gene expression information. Besides, the proposed method can also find subset of genes which may be responsible for Parkinson Disease. Chosen subset of genes causes the maximum generalization performance for PD classification problem. Proposed BCGA-ELM also produces a robust solution. In our experiments we executed BCGA-ELM twice started from randomly generated initial data and found same solution at the end. Detected set of 19 genes was also verified by SVM and PBL-McRBFN. Both methods caused maximum generalization performance.