Skip to content

f_temporal

Note Master 25-07-2026

The longest unbroken stretch of time the stay spent at the candidate, as a share of the whole stay:

\[f_\text{temporal}(E_k, C_i) = \frac{\max_{S \in S(E_k, C_i)} \sum_{p \in S} W_p}{\sum_{p \in C_i} W_p}\]

\(S\) is a run of points in time order that belong to candidate \(E_k\). \(W_p\) is the time weight from f_distribution. The score takes the longest run. A short jump outside the place does not end a run, as long as the total time outside stays within \(\Delta t_\text{gap} = 300\) seconds. Gaps longer than five minutes count as a new activity.

In the models: this score was removed after the first model. It was never significant, and it moved almost the same way as f_distribution (correlation r ≈ 0.88, VIF 7-8), so it added little new information. See Model 1: base.

Part of Framework overview.