Classification of Alzheimer related genes using LORENS with important and significant features

Prediction of Alzheimer case has been an important issue as it now becomes a degenerative disease. A lot of researches have been carried out to find out the best method for predicting the disease. One of the ways is by examining the gene expressions which lead to the classification whether the case is normal or Alzheimer. The gene expression is a macroarry data with the characteristic of high dimentionality leading to infeasibility of several classification methods to be applied. Feature selection has been a popular steps to analyze gene expression data. This paper investigates the performance of Logistic Regression Ensemble (LORENS) to classify Alzheimer case using important and significant features, selected by relative important criteria and logistic regression respectively. The results showed that predicting Alzheimer using LORENS with selected features (8 genes) improves the accuracy significantly compared to full features (20 genes), where the AUC reached 78.8% obtained from Cross Validation (CV) with 2 partitions.