Neural networks application in diagnosing alcohol addiction on the basis of the sleep analysis

The paper presents the outcomes of the neural networks (ANN) application in diagnosing alcohol addiction on the basis of the sleep analysis. Common sleep records, called polysomnograms of 172 people (men and women, half of them healthy, another half- addicted) were then adequately processed into parameters, which are characteristic for the patients and gathered in the data base. In our research including ANN systems, the data describing particular patients is optimized and simplified. The initial 26 parameters were reduced to the most essential 6, which are crucial in making diagnosis much more accurate than those given by a doctor (about 70%). Our neural systems verified by 20 fold cross validation allowed us to reach the accuracy in detecting addiction on the level of about 90 %. (Neural networks application in diagnosing alcohol addiction on the basis of the sleep analysis)