Elderly Movement Patterns and PageRank
Article:
https://www.hindawi.com/journals/ijta/2019/8612021/
Two researchers at Simon Frasier University in British Columbia, Canada, applied Google’s PageRank algorithm that we discussed in lecture in a novel way — using various sensors, they applied the algorithm to study anomalous movement patterns in the elderly. First, the researchers categorized areas of the elderly’s domicile, then mapped movements between rooms as directed edges, and finally ran the algorithm, normalized values, and compared normal room rankings with the elderly’s rankings.
The comparison between the researcher’s implementation and Google’s implementation is really quite creative. Rather than measuring the content and quality of a webpage, the researchers used a different metric — the probability that any elderly person would be in room Ri at time n+1 is modeled by the summation of the probability the elderly person is in a room that leads to Ri over the summation of the probability the elderly person is in a room that Ri leads to.
The result, a Markov matrix with probabilities that evolve over time n = 1, 2, 3 … has a steady state, which we’ve found in our usage of the PageRank algorithm as well. This state represents a very accurate probability that elderly people are in a room at any time, and the researchers used this information to find anomalous behaviors by comparing the resulting matrix with previously studied data.
The researchers interestingly found that the algorithm was very useful in detecting when the elderly visited the bathroom more frequently — thus resulting in ranking of the bathroom to be higher than normal in an anomalous pattern.
Overall I found this research to be a very intriguing application of Google’s PageRank algorithm, specifically because it models more than just anomalous locations (such as being in the bathroom more often) but because it models anomalous movements (such as going to the bathroom more frequently, from more locations), which has implications for senior health. This method of data collection and modeling seems promising, at least from my perspective. I think that in the same vein as this experiment, the PageRank algorithm could have applications in User Research (movement between UI components on a webpage, for example) to determine the most important elements in software, or even most important rollercoaster besides using just ridership metrics. I believe that how the researchers used PageRank to determine rooms’ “hub” and “authority” scores was creative and has promise for the health field.