File size: 8,699 Bytes
6951938
 
46d1433
 
 
 
 
 
 
 
6951938
3a3ccab
3b945df
 
 
 
 
 
 
3a3ccab
3b945df
 
 
46d1433
3b945df
 
 
 
 
 
 
 
 
 
 
46d1433
 
 
 
 
 
 
 
 
 
 
3b945df
3a3ccab
46d1433
3b945df
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3a3ccab
46d1433
3b945df
 
46d1433
 
 
 
 
 
 
 
 
 
 
3b945df
 
 
 
46d1433
 
 
 
 
 
 
3b945df
 
 
46d1433
 
 
 
 
 
 
 
 
 
 
 
3b945df
 
 
3a3ccab
3b945df
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
---
license: apache-2.0
task_categories:
- time-series-forecasting
tags:
- Spatial-Temporal
- Graph
- Logistic
size_categories:
- 10M<n<100M
---
# 1. About Dataset
**LaDe** is a  publicly available last-mile delivery dataset with  millions of packages from industry. 
It has three unique characteristics:  (1) Large-scale. It involves 10,677k packages of 21k couriers over 6 months of real-world operation. 
(2) Comprehensive information, it offers original package information, such as its location and time requirements, as well as task-event information, which records when and where the courier is while events such as task-accept and task-finish events happen. 
(3) Diversity: the dataset includes data from various scenarios, such as package pick-up and delivery, and from multiple cities, each with its unique spatio-temporal patterns due to their distinct characteristics such as populations.

If you use this dataset for your research, please cite this paper: {xxx}

# 2. Download
LaDe is composed of two subdatasets: i) [LaDe-D](https://huggingface.co/datasets/Cainiao-AI/LaDe-D), which comes from the package delivery scenario.
ii) [LaDe-P](https://huggingface.co/datasets/Cainiao-AI/LaDe-P), which comes from the package pickup scenario. To facilitate the utilization of the dataset, each sub-dataset is presented in CSV format.

LaDe can be used for research purposes. Before you download the dataset, please read these terms. And [Code link](https://github.com/wenhaomin/LaDe). Then put the data into "/data/raw/".  
The structure of "/data/raw/" should be like:  
```
* /data/raw/  
    * delivery    
        * delivery_sh.csv   
        * ...    
    * pickup  
        * pickup_sh.csv  
        * ...  
```

Each sub-dataset contains 5 csv files, with each representing the data from a specific city,  the detail of each city can be find in the following table.


|   City     |  Description                                                                                  |
|------------|----------------------------------------------------------------------------------------------|
| Shanghai   | One of the most prosperous cities in China, with a large number of orders per day.           |
| Hangzhou   | A big city with well-developed online e-commerce and a large number of orders per day.        |
| Chongqing  | A big city with complicated road conditions in China, with a large number of orders.          |
| Jilin      | A middle-size city in China, with a small number of orders each day.                          |
| Yantai     | A small city in China, with a small number of orders every day.                               |


# 3. Description
Below is the detailed field of each sub-dataset.
## 3.1 LaDe-P
| Data field                 | Description                                  | Unit/format  |
|----------------------------|----------------------------------------------|--------------|
| **Package information**    |                                              |              |
| package_id                 | Unique identifier of each package             | Id           |
| time_window_start          | Start of the required time window             | Time         |
| time_window_end            | End of the required time window               | Time         |
| **Stop information**       |                                              |              |
| lng/lat                    | Coordinates of each stop                      | Float        |
| city                       | City                                         | String       |
| region_id                  | Id of the Region                              | String       |
| aoi_id                     | Id of the AOI (Area of Interest)              | Id           |
| aoi_type                   | Type of the AOI                               | Categorical  |
| **Courier Information**    |                                              |              |
| courier_id                 | Id of the courier                             | Id           |
| **Task-event Information** |                                              |              |
| accept_time                | The time when the courier accepts the task    | Time         |
| accept_gps_time            | The time of the GPS point closest to accept time | Time       |
| accept_gps_lng/lat         | Coordinates when the courier accepts the task | Float        |
| pickup_time                | The time when the courier picks up the task   | Time         |
| pickup_gps_time            | The time of the GPS point closest to pickup_time | Time       |
| pickup_gps_lng/lat         | Coordinates when the courier picks up the task | Float        |
| **Context information**    |                                              |              |
| ds                         | The date of the package pickup                | Date         |


## 3.2 LaDe-D
| Data field            | Description                          | Unit/format   |
|-----------------------|--------------------------------------|---------------|
| **Package information**   |                                      |               |
| package_id            | Unique identifier of each package     | Id            |
| **Stop information**      |                                      |               |
| lng/lat               | Coordinates of each stop              | Float         |
| city                  | City                                 | String        |
| region_id             | Id of the region                      | Id            |
| aoi_id                | Id of the AOI                         | Id            |
| aoi_type              | Type of the AOI                       | Categorical   |
| **Courier Information**   |                                      |               |
| courier_id            | Id of the courier                     | Id            |
| **Task-event Information**|                                      |               |
| accept_time           | The time when the courier accepts the task | Time      |
| accept_gps_time       | The time of the GPS point whose time is the closest to accept time | Time |
| accept_gps_lng/accept_gps_lat | Coordinates when the courier accepts the task | Float |
| delivery_time         | The time when the courier finishes delivering the task | Time |
| delivery_gps_time     | The time of the GPS point whose time is the closest to the delivery time | Time |
| delivery_gps_lng/delivery_gps_lat | Coordinates when the courier finishes the task | Float |
| **Context information**  |                                      |               |
| ds                    | The date of the package delivery      | Date          |


# 4. Leaderboard
Blow shows the performance of different methods in Shanghai.
## 4.1 Route Prediction

Experimental results of route prediction. We use bold and underlined fonts to denote the best and runner-up model, respectively.

| Method       | HR@3         | KRC          | LSD         | ED          |
|--------------|--------------|--------------|-------------|-------------|
| TimeGreedy   | 57.65        | 31.81        | 5.54        | 2.15        |
| DistanceGreedy | 60.77        | 39.81        | 5.54        | 2.15        |
| OR-Tools     | 66.21        | 47.60        | 4.40        | 1.81        |
| LightGBM     | 73.76        | 55.71        | 3.01        | 1.84        |
| FDNET        | 73.27 ± 0.47 | 53.80 ± 0.58 | 3.30 ± 0.04 | 1.84 ± 0.01 |
| DeepRoute    | 74.68 ± 0.07 | 56.60 ± 0.16 | 2.98 ± 0.01 | 1.79 ± 0.01 |
| Graph2Route  | 74.84 ± 0.15 | 56.99 ± 0.52 | 2.86 ± 0.02 | 1.77 ± 0.01 |


## 4.2 Estimated Time of Arrival Prediction

| Method | MAE          | RMSE         | ACC@30      |
| ------ |--------------|--------------|-------------|
| LightGBM | 30.99        | 35.04        | 0.59        |
| SPEED | 23.75        | 27.86        | 0.73        |
| KNN | 36.00        | 31.89        | 0.58        |
| MLP | 21.54 ± 2.20 | 25.05 ± 2.46 | 0.79 ± 0.04 |
| FDNET | 18.47 ± 0.25 | 21.44 ± 0.28 | 0.84 ± 0.01 |


## 4.3 Spatio-temporal Graph Forecasting


| Method | MAE         | RMSE        |
|-------|-------------|-------------|
| HA    | 4.63        | 9.91        |
| DCRNN | 3.69 ± 0.09 | 7.08 ± 0.12 |
| STGCN | 3.04 ± 0.02 | 6.42 ± 0.05 |
| GWNET | 3.16 ± 0.06 | 6.56 ± 0.11 |
| ASTGCN | 3.12 ± 0.06 | 6.48 ± 0.14 |
| MTGNN | 3.13 ± 0.04 | 6.51 ± 0.13 |
| AGCRN  | 3.93 ± 0.03 | 7.99 ± 0.08 |
| STGNCDE  | 3.74 ± 0.15 | 7.27 ± 0.16 |



# 5. Citation
To cite this repository:

```shell
@software{pytorchgithub,
    author = {xx},
    title = {xx},
    url = {xx},
    version = {0.6.x},
    year = {2021},
}
```