s Manoglanistara - Emotional Wellness Phases Prediction of Adolescent Female Students by using Brain Waves

Background: Depression is the most underestimated and widespread health condition among people in developing countries. Depression levels among Indian population are rapidly increasing. It can be attributed to work pressure, social challenges, addiction to social media, adoption of the western culture and several other reasons. Indians’ depression levels are as high as 36 percent and shockingly this number is the highest in the world. Objective: What makes this even more alarming is the fact that WHO projects depression to be the second leading cause of disability worldwide by2020. Materials and Methods: This work focuses on Machine learning based Depression prediction by utilizing different brain wave frequency bands. It is carried out by askingthe universal standard Patient Health Questionnaire (PHQ.9) to subjects which are related to respective emotions. Neurosky’s Mind Wave Head kit is connected to the forehead (of subject) and 86 sample values are recorded. Total 85 Samples are trained, whereas 1 data is tested. Results: The MANOGLANISTARA- android App has been designed which sends the Emotional Wellness output (depressed/normal) to the subject via email. This provides the basis of analysis as to whether the subject is suffering from depression or not. Customization of the medication and treatment to such subjects can be initiated by the doctors. In this work, the MATLAB SVM based Depression prediction model is developed by evaluating the data built from Mindwave kit and standard PHQ.9 questionnaire. Work is also extended by using Orange Toolbox for classification of depressed/ normal subjects. Conclusion: In Orange toolbox, Prediction, ROC Analysis and Confusion Matrix are evaluated for different classifiers such as SVM, Naïve Bayes, Classification tree, Random forest and CN2 Rule Inducer. Accuracy, Precision, Sensitivity and Specificity is computed for all the abovementioned classifiers. CN2 Rule Inducer classifier gave higher accuracy of 0.9418, sensitivity 0.9778, Specificity 0.9736 and Precision 0.9778.