Name
Abstract | ||
|---|---|---|
Cycling provides various benefits to cyclists and cities. Nevertheless, the growth of cycling is still hindered by the lack of citywide information about perceived cycling safety. Providing cyclists with information about the safest routes could help increase cycling activity. In this paper, we aim to predict the perceived level of cycling safety for a trip (trip-PLOCS). We utilize LSTM-based architectures to incorporate the sequential information of segments in a trip, and predict its cycling safety. Our proposed method can achieve up to 76% F1 micro (65% F1 macro) score, 10% (19%) better than the state-of-the-art baseline. Finally, we use SHAP to extract insights about trip-PLOCS, showing that social features contribute to perceived danger while cycling facilities contributes to the perceived safety.
|
Year | DOI | Venue |
|---|---|---|
2019 | 10.1145/3347146.3359092 | SIGSPATIAL/GIS |
Keywords | Field | DocType |
cycling safety prediction, spatio-temporal analysis, open data | Computer science,Transport engineering,Cycling,Artificial intelligence,TRIPS architecture,Machine learning | Conference |
ISBN | Citations | PageRank |
978-1-4503-6909-1 | 0 | 0.34 |
References | Authors | |
0 | 3 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Jiahui Wu | 1 | 194 | 14.61 |
| Lingzi Hong | 2 | 18 | 3.35 |
| Vanessa Frías-Martínez | 3 | 107 | 10.32 |