Fig 1.
Study area (thick dashed line) context map in Greater Melbourne.
Orange dots indicate sensor locations. Inset rectangle delineates the area covered in Figs 4 and 5. Wikimedia base map accessed under Open Data Commons Open Database License (ODbL), created by OpenStreetMap contributors.
Table 1.
Change in built environment in Melbourne CBD, 2014–2018.
(Source: City of Melbourne 2021).
Table 2.
Descriptive statistics of June weekday pedestrian counts from 32 sensors that were consistently active from 2014–2018 during AM, lunch, and PM periods.
(Source: City of Melbourne 2020).
Fig 2.
Betweenness results for street segments for a single origin-destination pair, where the origin has a weight of “100” (of which only 20.2 are routed due to a distance decay effect) along all routes that are up to 15% longer than the shortest path.
Fig 3.
Potential 36 origin-destination pairs for pedestrian trips in Melbourne.
Gray highlights include the ten O-D pairs for which trips were modeled. Weights describe which attribute was used as origin or destination weight in the analysis.
Table 3.
Goodness of fit results of MLM and the chosen SVR model for 10 types of pedestrian flows combined with weather and day-of-week variables, with sensor-level dummies.
Left: calibration results on June 2014 data. Right: prediction results on June 2015 data.
Table 4.
Evaluating impact of built-environment measures on prediction with sensor dummies (left) and without (right).
Prediction metrics for 2015 using a model calibrated on 2014 are presented.
Fig 4.
Predicted pedestrian flows during a typical workday AM peak period (8.00–9.00AM) in June 2015, based on June 2014 calibration.
Fig 5.
Predicted change in pedestrian flows during a typical workday AM peak period (8.00–9.00AM) in between June 2014 and 2015.
Fig 6.
Effect of pedestrian flow measures on model predictions without sensor-level dummies during AM peak, Lunch peak, and PM peak.
Table 5.
Examining stability of model prediction performance without sensor-level dummies over time (using RMSE).
Fig 7.
Stability of feature importance from 2014–2019 for AM, lunch and PM peak periods (without sensor-level dummies).