Mixed dementia is characterized with the presence of both vascular and Alzheimer's-related changes and is currently the largest cause of dementia in older age. Engagement in physical activity has been found to be an effective way to lower dementia risk, and yet the percentage of adults who consistently remain physically active remains low. Functional MRI provides insights into the structures and neural mappings of the brain, insights that can be used to predict the neural and cognitive basis for an individual's actions and habits. More specifically, resting-state functional connectivity (RSFC) can predict behavior by taking into account the brain's connections and data on an individual's behaviors. This study examines the interplay between social, cognitive, and behavioral factors (sociodemographic factors, behavioral traits, cognitive abilities, social factors, the physical environment, and baseline RSFC values) in determining what motivates previously-inactive older adults to start exercising after receiving a new cardiovascular diagnosis. The data used in this study came from 295 physically inactive older adults who received a new cardiovascular diagnosis in the time period between an initial visit and a follow-up visit that occurred 4 years later. "Physically inactive" was defined as getting less than the World Health Organization recommendation of 150 minutes of moderate-to-vigorous physical activity (MVPA) per week. MVPA values were both subjective and objective - subjective meant participants self-reported at the initial and follow-up visit while objective meant that some participants were randomly selected to wear an accelerometer on their wrist for a week to track their physical activity level. Tools such as the Townsend deprivation index, cognitive tests, self-reported assessments, ratings on numerical scales, and Euclidean geometry were all ways to measure the participants' socioeconomic status, cognitive ability, mental health, strength of their social support systems, and proximity to any green spaces. Functional MRIs were taken and postprocessed to remove any additional randomness in the neural network that came from cerebrospinal fluid and white matter and any instances of subject head motion, as head motion can impact neural connectivity. 400 RSFC features were selected by analyzing prior research to determine the features strongly correlated with cognitive and physical behavior. The support vector machine (SVM) regression algorithm was used to predict future succesful physical activity from the objective MVPA values; three individual models were trained with these algorithm, one model strictly relying on the measured social/behavioral/cognitive factors mentioned earlier, one model strictly relying on the neural imaging as determined by the functional MRIs, and the last model integrating both the measured factors and the neural imaging. After receiving a new cardiovascular diagnosis, participants demonstrated a signficant average increase of 7.52 minutes per week. There was a non-significant declining trend in cognitive ability between the initial and follow-up visit (the P value did not reach the statistical significance of 0.05). However, those with a higher self-reported MVPA at the follow-up visit compared to the initial visit demonstrated improved working memory over time, with a statistically significant P value of 0.02. The SVM model that was trained strictly on the measured factors was unable to predict future MVPA values at a statistically significant level (P = 0.056), while both the SVM model trained strictly on neural imaging and the SVM model trained on the measured factors in tandem with neural imaging were able to predict future physical activity at a statistically significant rate (P = 0.002 and 0.0008, respectively). Future physical activity was also able to be predicted from the subjective MVPA values via two models - one model only looked at the features generated from RSFC data and the other model integrated RSFC data with data from the other measured cognitive, behavioral and social factors. Both models were predicting physical activity at statistically significant levels, with P values of 0.004 and 0.001, respectively. The best-performing integrated model (i.e. model that combined measured factors and RSFC data) indicated that percentage of green space in a neighborhood, frequency of visits from loved ones, and retirement status was positively correlated with MVPA change. The RSFC features in the best-performing integrated model for predicting a change in future physical activity were all located in the left hemisphere and spanned multiple locations and networks, such as the temporal lobe and medial prefrontal cortex or the lateral prefrontal cortex.