Classifying smoking urges via machine learning

摘要: We examine different machine learning approaches to classify smoking urges.We use situational features associated with urges smoke during a quit attempt.Feature selection algorithms make possible obtain high classification rates.The tree method provides the best results. Background and objectiveSmoking is largest preventable cause of death diseases in developed world, advances modern electronics can help us deliver real-time intervention smokers novel ways. In this paper, we having or not attempt order accurately high-urge states. MethodsTo test our approaches, specifically, Bayes, discriminant analysis decision methods, used dataset collected from over 300 participants who had initiated attempt. The three are evaluated observing sensitivity, specificity, accuracy precision. ResultsThe outcome showed that based on feature it rates only few selected entire dataset. outperformed naive Bayes an classifications up 86%. These numbers suggest may be suitable approach deal cessation matters, predict urges, outlining potential for mobile health applications. ConclusionsIn conclusion, classifiers identify situations, search classifier parameters significantly improves algorithms' performance. addition, study also supports usefulness new technologies improving effect interventions, management time patients by therapists, thus optimization available care resources. Future studies should focus providing more adaptive personalized support people really need it, minimum amount developing expert systems capable delivering interventions.