Abstracts Psychology

Design of a system for the detection of hidden patterns in the emotional impact and coping strategies on oncology patients using supervised learning techniques

by Natalia Rodríguez Núñez-Milara

The Institute of Psychology and Emotion, is a psychology company linked to the world of research that has been conducting research projects for over 20 years with the university in the field of health psychology". After his team carried out several works in the field of cancer, the need was detected to be able to create a screening questionnaire that would allow an evaluation of various psychological variables such as anxiety, depression, avoidant coping or helpless coping, in a preventive manner. Within this context, this project arose from the need to be able to analyze in a comprehensive manner, all the data to be collected for the elaboration of the screening. In addition, since the patients were going to be able to answer several questionnaires beforehand, it was proposed to be able to acquire extra information to be able to evaluate it in this report. This information includes data on patients' life habits, socioeconomic data, data on anxiety and depression, data such as age, or protocols on communication between doctor and patient in various circumstances. In this project we have proceeded to perform an exhaustive analysis of all the data collected, as well as the possible implications of various variables in the psychological affectation of cancer patients or in the course of their disease. In addition, this project explores the possibility of using supervised learning techniques to obtain better results from the analysis of the questionnaires, as well as the development of a preventive classification tool to evaluate the degree of certain psychological variables, depending on the emotional state. Different architectures of end-to-end systems and Machine Learning models for classification will be studied (8 different models) as an approach to predict the results within a patient evaluation. In addition, data will be explored to find hidden patterns among the patients participating in the research. The results of the study show that there are classifiers that can have great results by labeling patients into 5 groups in the evaluation process depending on the number of psicologycal variables that have high (0,1,2,3 or 4).