Deploying PsyKohonen Unsupervised Machine Learning Algorithm for Diagnosing Personality Disorders in three Clusters as per DSM-5

Abstract

The transferences of the heritable information are unconsciously influencing human lives, relationships and social bonding one forges into. This means that next to the self-determination, free will, and the intentions by which one designs his / her life, there is a considerable genetic programming from the ancestors with a person’s emotions, experiences, beliefs, and shortcomings he / she might have. Besides that genetic predisposition, sufficing impact on a person’s life is also influenced by the environmental conditioning and demands. When the environmental or societal demands are not met or fulfilled, and such deficiency is affecting the holistic functioning of a person in different aspects of life, then he / she is attributed to have some personality limitation and eventually a disorder when gauged on or diagnosed with DSM-5 metrics.  Benefitted through the professional experience gained in the capacity of Psychological Counselor in one of the oldest Anglo Indian Higher Secondary Schools in India, the first author of this research paper has made use of the dataset from 100 respondents for the Personality Inventory Questionnaire that he produced. Such personality inventory questionnaire contains five close-ended options for each of the fifty questions. Responses to those questions are used to set weights and threshold. This research paper deals with deploying an innovative unsupervised machine learning algorithm called PsyKohonen, a derivative improvement of the Kohonen SOM Algorithm; for diagnosing the personality disorders that may fall in one of the three personality disorder clusters as described in DSM-5. PsyKohonen is a very modernistic approach to diagnosing personality disorders that draws inspirations from the original Kohonen SOM Algorithm, by incorporating winner take-all, and lateral inhibition strategies. If a subject’s responses imply the maximum total in any one of the three personality disorder clusters, by considering dataset comprising of five questions for each personality disorder, then the subject could be suffering from that particular disorder represented by the corresponding cluster, namely Cluster A – indicating eccentric disorders, Cluster B – revealing erratic disorders, and Cluster C – denoting anxious disorders. PsyKohonen machine learning algorithm is deployed in Python 3 in Google Colab IDE. NumPy, Scikitlearn, Matplotlib PyPlot libraries were used for the extraction of results and visualization of the PsyKohonen’s functionalities. The last segment in this paper, sheds light on the horizontal bar plot, and pie chart indicating the prediction of diagnosis produced by the PsyKohonen Algorithm, and the Interpretation for such visualizations.