Improved Random Forest Algorithm to Classify Methicillin-Resistant and Methicillin-Susceptible Staphylococcus Aureus on Mass Spectra.

Mass spectrometry (MS) method has been one of the most popular subjects in the field of microbial identification by the reason of its rapid identification and variety application. However, the biomarker of Methicillin-resistant and methicillin-susceptible Staphylococcus aureus is out of measuring range by mass spectrometer, which leads to hard to classify for the molecules by those instruments. In this paper, to classify the molecules based on the MS in the small measurable range, we propose a reducing dimensions algorithm to deal with the heterogeneity of variables. With the preprocessed data, a random forest (RF) is used to identify the methicillin-susceptible S. aureus (MSSA) and methicillin-resistant S. aureus (MRSA). Experiments verify the accuracy of proposed methods. The result shows that the accuracy, recall, false positive rate and precision of proposed method are more than 90 percent. For medical institutions, the method which we proposed could identify MRSA from MSSA rapidly and save-costing on mass spectrometry data-set.