Title:
LSTM-BASED DIDI TAXI ORDER DEMAND PREDICTION METHOD AND APPARATUS
Document Type and Number:
WIPO Patent Application WO/2022/174434
Kind Code:
A1
Abstract:
An LSTM-based Didi taxi order demand prediction method and apparatus. The method comprises the following steps: screening and cleaning acquired GPS order data to obtain preprocessed data, and matching the preprocessed data with an actual map road network (101); performing clustering analysis on the preprocessed data to obtain regional data clusters, and dividing the actual map road network into several sub-regions according to the regional data clusters (102); according to the regional data clusters in the sub-regions, training an order prediction model on the basis of a long short-term memory network (103); predicting, by means of the order prediction model, the number of starting point orders and the number of ending point orders in each time period of each sub-region, and searching for hot spot regions (104); and calculating an order difference between the hot spot regions, and determining taxi demands in the sub-regions in the time period according to the order difference (105). The method is beneficial for improving the economic benefits of Didi taxi drivers and improving the riding experience of passengers.
More Like This:
Inventors:
LI YING (CN)
YANG XIAOMENG (CN)
ZHU DONGYANG (CN)
YANG RUNJIA (CN)
TIAN RUXIA (CN)
HE WEI (CN)
GAI TENGFEI (CN)
YANG XIAOMENG (CN)
ZHU DONGYANG (CN)
YANG RUNJIA (CN)
TIAN RUXIA (CN)
HE WEI (CN)
GAI TENGFEI (CN)
Application Number:
PCT/CN2021/077139
Publication Date:
August 25, 2022
Filing Date:
February 22, 2021
Export Citation:
Assignee:
CHANGAN UNIV (CN)
International Classes:
G06K9/62
Foreign References:
| CN111753910A | 2020-10-09 | |||
| CN107392389A | 2017-11-24 | |||
| CN109583611A | 2019-04-05 | |||
| CN112150207A | 2020-12-29 | |||
| US20180018572A1 | 2018-01-18 |
Attorney, Agent or Firm:
BEIJING GLOBE-LAW LAW FIRM (CN)
Download PDF: