Good Machine Learning Path Opportunities

Question

I study ML problems (mainly density estimation and anomaly detection) with coordinate paths (GPS). Besides the coordinates themselves and the delta (changes between neighboring coordinate points) and polar coordinates, what are some other good features? What features make intuitive attributes such as straightforwardness, seductiveness, smoothness and concreteness?

Answer 1

For straightness / curvature, you can calculate the approximate first derivative of the curve, for smoothness, the second and highest derivatives. If by closure you mean a tendency to return to places several times, you could, for example, calculate how many segments intersect each other.