Artificial intelligence based classification of specific neuropsychological disorders

Artificial Intelligence technologies are used in many areas, including the medical field. One of these areas is the detection, diagnosis and, if necessary, treatment of neuropsychological diseases. These diseases are disorders that we frequently encounter in our daily lives. There are many disorders within the scope of neuropsychology. If these disorders are not detected, both the individual's life may be in danger and it causes a difficult process for the environment. This study has the feature of being the first thesis study in the literature that classifies many neuropsychological disorders based on artificial intelligence. Within the scope of this study, five different data sets including anxiety disorder, schizophrenia disorder, autism spectrum disorder, depression and dementia disorders were used. Studies were conducted on the data sets by using data pre-processing, hyperparameter optimization and artificial intelligence models. XGBoost, LightGBM, Random Forest, Support Vector Machines and k- Nearest Neighbor were used as machine learning models and comparisons were made on accuracy metrics. In addition, before and after comparisons were made on machine learning models using hyperparameter optimization. For data set 1, LightGBM technique had the highest success rate with 96% accuracy, while after hyperparameter optimization, SVM came to the forefront with 97% accuracy and gave the highest accuracy rate. For data set 3, XGBoost and LightGBM techniques achieved a successful result with 98% accuracy, while this rate did not change after hyperparameter optimization and XGBoost gave the most successful result. As deep learning techniques, Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM) and Gated Repetition Unit (GRU) models were designed in accordance with the data sets. For data set 2, LSTM model gave 83% accuracy rate. For data set 4, GRU model gave the highest accuracy rate with 93%. For data set 5, LSTM and GRU gave a high result with 99% accuracy rate. Deep learning models were not only compared with their accuracy rates but also with their learning curves. This study clearly shows that it is possible to classify neuropsychological disorders with artificial intelligence.