Table 1 Summary of AI methods integrated into the time-activity model
AI method | R implementation | Outputs |
|---|---|---|
Home-range method that combines geometric and probabilistic estimators | Time Local Convex Hull (T-LoCoH) [34] | Polygon (hull) geometry gives information on directional movement vs. static clusters (Step 2). Visitation rate and duration of visit enable classifications based on behavioural patterns of the individual (Step 3). |
Trajectory analysis | Adehabitat LT [37] | Segmentation of movement with the Lavielle method [57] |
Predictor selection for Random Forest (RF) classification with three-step elimination process based on data-driven thresholds for high dimensional datasets | VSURF [38] | Predictor variables for RF models collected with the PAM (movement, noise, GPS information) and baseline questionnaire (common modes of transport), and extracted from spatial analysis |
RF classification of the mode of transport with the 10-fold evaluation method | RandomForest [65] | Probabilistic classification for each mode of transport |