| dc.description.abstract | Staphylococcus aureus is a Gram-positive globular bacterium that is a flora common to the epidermis, but often causes opportunistic infections and is one of the common infectious strains. Among them, methicillin-resistant Staphylococcus aureus, known as ＂superbug＂, is a common infectious strain and is resistant to most penicillin antibiotics. Traditional identification of drug resistance often takes about three days, which is usually dependent on the physician′s empirical therapy. If the drug resistance is identified early and treated with appropriate drugs, it will greatly improve the effect of treatment. The purpose of this study was to use the mass spectrum of the strain obtained by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry, and combine the machine learning to quickly classify the resistance of this strain to oxacillin, clindamycin and erythromycin. This study uses decision trees, random forests and support vector machine classifiers to make prediction models, and finds that using random forest classifiers with sequential forward selection to select features can have the highest accuracy. The accuracy of predicting oxacillin, clindamycin and erythromycin resistance on independent test set were 80.56%, 82.42% and 74.71%, respectively. The results predicted by our model can provide a timely reference for clinicians, and provide more appropriate care for patients. | en_US |