Machine learning methods are utilized in a wide variety of fields, such as healthcare, finance, and marketing, and have shown great success in them. They allow automating analytical model construction in order to draw quantitative conclusions from data. In this research we examine the performance of various machine learning models, and specifically classifiers, on cognitive data, in both typical and clinical populations.
In Study 1, focusing on typical (healthy) population, we use machine learning models to predict one’s level of intelligence, specifically fluid intelligence, based on various cognitive and behavioral measures, such as personality and behavior. We find that intelligence can be predicted quite well using only a small subset of features.
In Study 2, focusing on clinical populations, we use machine learning models in conjuncture with a multidimensional representation of the mental lexicon (via a cognitive multiplex network model) to predict whether a patient has dementia and if so, the specific type of dementia they have, based on a simple semantic fluency task. We find that these models can predict almost perfectly whether a patient has dementia, and considerably well the specific type of dementia.
Overall, in both studies we demonstrate the significance of applying machine learning algorithms to study cognition, in both typical and clinical populations.