1 Introduction
-
‘travelers’: users that move around extensively, visit different areas with distinct, salient characteristics and do not stay long at a specific location.
-
‘local residents’: users that move in a more constrained area, e.g. a city, usually cover only short distances and revisit many of their locations.
2 Related work
2.1 The value of EHMD
2.2 How to process EHMD
2.3 Data mining with movement data
2.4 Travelers and locals: characterization of users by movement behavior
3 Methodology
3.1 Data
3.2 Overview
3.3 Data cleaning
3.4 Feature extraction
Feature name | Clarification and description | Feature type | |
---|---|---|---|
1 | Active days | The number of days the user has used the app | Temporal Activity |
2 | Consecutive days | The highest number of consecutive days the user has used the app | Temporal Activity |
3 | Weekdays | The number of distinct weekdays the user has used the app | Temporal Activity |
4 | Period | The number of days between the first and last usage | Temporal Activity |
5 | Total time | The total amount of time the app has been running [s] | Temporal Activity |
6 | Variation of time stamp | The standard deviation of all time stamps [s] | Temporal Activity |
7 | Total distance | The total distance the user has covered [m] | Spatial Distance |
8 | Maximum distance | The distance between two most distant points a user has visited [m] | Spatial Distance |
9 | Daily distance | The average distance covered in a day [m] | Spatial Distance |
10 | Variation of daily distance | The standard deviation of the average distances per day [m] | Spatial Distance |
11 | Daily centroid distance | The average distance of two consecutive1 daily centroids [m] The centroid is the centroid of the daily concave hull | Spatial Distance |
12 | Variation of daily centroid distance | The standard deviation of the distance between two consecutive daily centroids [m] | Spatial Distance |
13 | Distance to centroid | The average distance between the daily centroid and the overall centroid [m] | Spatial Distance |
14 | Variation of distance to centroid | The standard deviation of the distance between the daily centroid and the overall centroid [m] | Spatial Distance |
15 | Average step length | The average distance covered in a move segment2 [m] | Spatial Distance |
16 | Standard deviation of step length | The standard deviation of the distances covered in move segments [m] | Spatial Distance |
17 | Area | The total area the user has covered [m2]3 | Spatial Area |
18 | Circumference | The circumference of the total area the user has covered [m] | Spatial Area |
19 | Complexity | The complexity of the total area (area/circumference) | Spatial Area |
20 | Compactness | The compactness of the total area [4∗area/π∗maximum distance squared] | Spatial Area |
21 | Daily area | The average area of the daily areas covered [m2] | Spatial Area |
22 | Variation of daily area | The standard deviation of the daily areas covered [m2] | Spatial Variability |
23 | Overlap | The average percent of overlap of two consecutive daily areas covered [%] | Spatial Variability |
24 | Variation of overlap | The standard deviation of the percentage of overlap of two consecutive daily areas covered [%] | Spatial Variability |
25 | Spatial clusters | The number of clusters of start, stop or end points4 | Spatial Variability |
26 | Number of moves | The absolute number of move segments | Spatio-Temporal Dynamics |
27 | Average speed | The average speed in the move segments [m/s] | Spatio-Temporal Dynamics |
28 | Standard deviation of speed | The standard deviation of the speed in the move segments [m/s] | Spatio-Temporal Dynamics |
29 | Number of stops | The total number of stops | Spatio-Temporal Dynamics |
30 | Total stop duration | The total duration of all stops [s] | Spatio-Temporal Dynamics |
31 | Stop duration | The average duration of a stop [s] | Spatio-Temporal Dynamics |
32 | Variation of stop duration | The standard deviation of the stops [s] | Spatio-Temporal Dynamics |
3.5 Iterative feature reduction and clustering
3.6 Finding an optimal clustering approach
3.7 Interpretation
-
Cluster 1 (4451 unique users): Cluster members are inactive. They turn on the app only sporadically. Thus, their recorded movement is episodic and erratic. Moreover, users in Cluster 1 move slowly and stop frequently.
-
Cluster 2 (3343 unique users): Cluster members are active, cover large distances and visit many different locations in Australia.
-
Cluster 3 (5861 unique users): Cluster members are highly active, but only roam in relatively confined areas.
-
Cluster 4 (1079 unique users): Similarly to Cluster 1, cluster members are inactive and only use the app sporadically. However, in contrast to Cluster 1, they move fast and stop only infrequently.
-
Cluster 5 (4372 unique users): Cluster members are active. On the one hand they show a behavior that lies between that found in Cluster 2 and Cluster 3, for example with respect to the distance travelled, the total area visited, or the compactness. On the other hand, they behave similarly to Clusters 2 and 3, e.g. with respect to the number of spatial clusters5 (Cluster 2), and with respect to overlap (Cluster 3).
3.7.1 Travelers
-
Users are familiar with some locations on their overall route, but not with all. Thus, they turn on the navigation app only sporadically.
-
Users visit different regions in Australia, which they reach by other means of transport (e.g. planes). For these travels they do not need the navigation app. This behavior could be typical for tourists (in particular, domestic tourists but also business travelers). The large distances between major cities in Australia cause this to be a typical tourist behavior.
3.7.2 Locals
4 Spatio-temporal behavior of user types
4.1 Aggregated temporal patterns
Travelers | Locals | ||||||
---|---|---|---|---|---|---|---|
Mean | Std. dev. | CV | Mean | Std. dev. | CV | ||
Sydney | Hourly | 303.06 | 202.30 | 0.67 | 1741.80 | 1035.43 | 0.59 |
Daily | 3636.71 | 238.96 | 0.06 | 20,901.57 | 2607.68 | 0.12 | |
Melbourne | Hourly | 323.35 | 227.23 | 0.70 | 1895.36 | 1089.02 | 0.57 |
Daily | 3880.14 | 284.43 | 0.07 | 22,744.29 | 2792.68 | 0.12 |