Research Article
Recurrent Neural-Based Vehicle Demand Forecasting and Relocation Optimization for Car-Sharing System: A Real Use Case in Thailand
Table 1
A comparison of usage logs among (1) unprocessed data, (2) cleaned data by experts, and (3) cleaned data by our algorithm.
| | Experiment | Unprocessed | Cleaned by experts | Cleaned by system |
| | September 2019 | | | | | Failed usage (no vehicle case) | 592 | 150 (−442) | 297 (−295) | | Failed usage (no station case) | 545 | 174 (−371) | 263 (−282) | | Total failed usage | 1,137 | 324 (−813) | 560 (−577) |
| | October 2019 | | | | | Failed usage (no vehicle case) | 715 | 130 (−585) | 322 (−393) | | Failed usage (no station case) | 492 | 149 (−343) | 227 (−265) | | Total failed usage | 1,207 | 279 (−928) | 549 (−658) |
|
|
The number in the parentheses refers to the number of fake usages that can be reduced.
|
