Description
Data-driven methodologies to predict and potentially prevent opioid use disorder and overdose have risen alongside the opioid crisis. We aimed to clarify this literature by directly comparing the effectiveness of four common machine learning algorithms using data from the Military Health System. Support vector machine was the most accurate, yielded the highest c-statistic, and was second in predicting all other analyses, indicating SVM is likely the best all-purpose algorithm for this domain.
Describe the new knowledge and additional skills the participant will gain after attending your presentation.: They will understand four common types of machine learning algorithms and why support vector machine was the best at predicting a high number of opioid prescriptions in the Military Health System.
Authors:
Emily Stone (Presenter)
Defense Health Agency
Presentation Materials: