Data mining and machine learning for identification of risk factors and prediction of cognitive changes among aging populations

Cognitive decline is a common consequence of aging, with dementia at the extreme end of this process. The decline in cognition may decrease the ability and efficiency of performing daily living activities among older adults. Unfortunately, existing pharmacological treatments are not effective at delaying the incidence of dementia and cognitive impairment. As such, many medical recommendations are focused on preventative measures (i.e., lifestyle activities, social engagement, physical activity, and proper diet) to maintain cognitive health. Although the results of the previous studies in this area are promising, there are yet unanswered questions that restrict the practical applications and recommendations of the interventions and their impact on cognition. This thesis research investigates the application of data mining methods to answer some of the yet unanswered questions. Accordingly, this thesis first aims to investigate the impact of engagement in different intensities and frequencies of physical activity on two domains of cognitive function. We seek to test the hypothesis that engaging in a physically active lifestyle leads to relatively preserved cognitive health during aging. The findings of the study assist communities to promote healthy cognitive aging among older populations by implementing new policies and providing recommendations about the details of engaging in optimal physical activity in terms of intensity and frequency. Second, we aim to focus on the impact of cognitive reserve on cross-sectional cognitive function, short-term and long-term rates of cognitive changes over 2 years and 10 years of follow-up. Our objective is to attempt to improve the limitations of previous studies in terms of study design, intervention characteristics, and methodological issues. Our use of data mining approaches and appropriate study design models assist in controlling the impact of confounding factors and moving forward towards investigating the causal relationship rather than correlational association. The results of this study contribute to establishing interventions to be developed during the aging process to delay cognitive decline. Lastly, we aim to investigate the possibility of implementing a model to predict future cognitive changes with the combination of categorical and continuous data from multiple domains such as sociodemographic, health, psychology, and cognition, simultaneously. We also use a machine learning-based framework to identify the most important predictors of future cognitive changes. Incorporation of the findings of the study in public health policies assists in improving the counseling of older adults and caregivers and developing the plan of cognitive care and effective interventions to develop healthy aging.