Detecting Comorbidity Using Machine Learning

Comorbidity between mental disorders is a topic of great interest as professionals in the field have been recognizing emerging patterns between a variety of disorders. This research experimented with neural networks to find if machine learning could be as accurate as professional diagnoses and concluded with two neural network models that provided the highest accuracy for diagnosing comorbidity. The training data used was retrieved from ”Psychiatric Comorbidity in Patients from the Addictive Disorders Assistance Units of Galicia: The COPSIAD Study,” (Pereiro et al., PLoS One 8(6):66451, 2013) as it pertains to comorbidity and doing statistical calculations on its real-world data and visualizing it for use as training data in the machine. The result concluded with the two models predicting certain relationships very well while others did not. More data will be needed to improve the accuracy of a professional diagnosis but the possibility of machine learning being used as a method of diagnosis may be possible in the future, as exhibited in this research.