San Francisco has provided data on pedestrian counts at street intersections. According to the map, we see that most pedestrians were registered in the North-Eastern region of the city.
But what can we do with this data? These are only counts with associated geocoordinates. The question we might ask ourselves is how people come from one point to another. Were they walking, driving a car or using other means of transportation? If they were walking from one point to another, they would increase the counts at neighboring intersections and if they were driving, they would likely appear in another statistic. If they were walking to reach a destination that would be most beneficial to them, then we could view their walks similar to the way we optimize a given function. If we remember gradient descent, we know that the oscillation in initial phases is stronger than the one we obtain once we approach the optimum. The extent to which something similar happens while people are walking would be equally interesting to observe.
We could order the counts and connect the intersections with similar counts. If we had points A, B and C, and counts(A) < counts(B) < counts(C), then we would connect A with B and B with C, but not A and C. We connect in the order of the counts to get the following map.
You can hover on a point to see which streets are intersecting there. We notice two clusters of points and relatively many line crossings between them, which may indicate more frequent traveling in relation to walking since if these line crossings were reached by walking, the pedestrian counts would have appeared on the map as well. Starting from the smallest point at the upper right, we see that the first connections are mainly between points within the two separate clusters, so the path lengths are correspondingly long. But once a cluster is reached, we see that the paths become much shorter and more numerous. They eventually lead to a place that people were most often gravitating to.