Candidate selection

Note Master 25-07-2026

How the method decides which places to consider for one stay. Draw a circle with radius R_buffer around every GPS point of the stay. Every place that touches at least one circle becomes a candidate:

\[\text{Candidates}(C_i) = \{E_k \mid \min_{p \in C_i} d(p, E_k) \leq R_\text{buffer}\}\]

In words: place \(E_k\) is a candidate if its distance \(d\) to at least one point \(p\) of stay \(C_i\) is at most \(R_\text{buffer}\).

A circle is used instead of "the point is inside the place", because GPS error can put a recorded point outside the place where the person really was. f_distribution uses the same radius on purpose.

R_buffer = 10 m was chosen over 15 m: R_buffer: 10 m vs 15 m. Because each point has its own circle, a stay with more points covers a larger area: Downsampling shrinks the candidate search area.

Part of Framework overview.

Linked from Assignment framework · Downsampling shrinks the candidate search area · Framework overview · R_buffer: 10 m vs 15 m