Predicting asthma outcome using partial least square regression and artificial neural networks
Advances in Artificial Intelligence
| Hi-index | 0.00 |
Motivation: Machine learning in bioinformatic sheds light on the traditional biography research. Through the prediction of functional genes from amino sequence information, the experimental cost for new gene finding could be reduced. Results: We propose an effective machine-learning approach based on artificial neural networks (ANN), to assess the chance of a protein in rice to be disease resistant or not. Through feature reduction, 30 important feature correlated to disease-resistance are discovered. The feature selection approach results in 92.86% reduction of number of features. Afterwards, we construct a feature reduced classifier. The accuracy of the new classifier achieves 100% in resubstitution test and 72.13% in Jackknife test, and the Matthews correlation coefficient achieves 0.4419. Eventually, top 10 possible Xoo-resistant genes are found.