Capstone Projects for Professional Certificate in Data Science from Harvard
Falls among the elderly population is a serious health concern. In fact, the CDC reports that the mortality date of falls for people aged 65 and older rose by 30% between 2007 and 2016. Perhaps just as concerning is that while about 20% of geriatric falls results in a serious injury, less than half of elders who experience falls notify their doctors. The advent of wearbale smart devices, however, offers a chance to change this. If data from the accelerometers and vitals sensors in smart watches can be analyzed to distinguish falls from other daily activities, it would allow health care providers or family members to recevie instant alerts of an elderly person’s fall. In turn, this would enable a more rapid response to falls that necessitate medical attention.
This project uses the famous MovieLens dataset used to simulate the data from the Netflix Recommendation Algorithm challenge. In order to win the $1 million dollar prize, teams had to reduce their RMSE to about 0.857. The algorithm built in my project was able to provide RMSEs well below that, with a final RMSE score on the test set of 0.825.