Auto-analyzing spatial relationships in multi-scale spatial datasets for spatio-temporal prediction

Patent metadata

Field	Value
Publication number	US-10600007-B2
Application number	US-201414450712-A
Country	US
Kind code	B2
Filing date	Aug 4, 2014
Priority date	Aug 4, 2014
Publication date	Mar 24, 2020
Grant date	Mar 24, 2020

Abstract

A method and system to perform spatio-temporal prediction are described. The method includes obtaining, based on communication with one or more sources, multi-scale spatial datasets, each of the multi-scale spatial datasets providing a type of information at a corresponding granularity, at least two of the multi-scale spatial datasets providing at least two types of information at different corresponding granularities. The method also includes generating new features for each of the multi-scale spatial datasets, the new features being based on features of each of the multi-scale spatial datasets and spatial relationships between and within the multi-scale spatial datasets. The method further includes selecting, using the processor, features of interest from among the new features, training a predictive model based on the features of interest, and predicting an event based on the predictive model.

First claim

What is claimed is:

1. A method of performing spatio-temporal prediction, the method comprising:
obtaining, based on communication with one or more sources, multi-scale spatial datasets, each of the multi-scale spatial datasets providing a type of information at a corresponding granularity, at least two of the multi-scale spatial datasets providing at least two different types of information at different corresponding granularities, wherein each granularity defines a minimum area to which the corresponding multi-scale spatial dataset corresponds;
generating, using a processor, spatial relationships both between the multi-scale spatial datasets at the different corresponding granularities and within the multi-scale spatial datasets;
generating, using the processor, features from each of the multi-scale spatial datasets, wherein each feature of each of the multi-scale spatial datasets is a unit of the multi-scale spatial dataset;
generating, using the processor, new features for each of the multi-scale spatial datasets, the new features being based on the features of each of the multi-scale spatial datasets and the spatial relationships between and within the multi-scale spatial datasets;
selecting, using the processor, features of interest from among the new features;
training a predictive model based on the features of interest; and
predicting an event based on the predictive model.

2. A method according to claim 1 , wherein at least two of the multi-scale spatial datasets are in at least one pairing.

3. The method according to claim 2 , further comprising developing the spatial relationship between each pairing of the multi-scale spatial datasets as a pair of distributing matrices.

4. The method according to claim 3 , wherein the developing the spatial relationship between a pairing of a first dataset and a second dataset among the multi-scale spatial datasets, the first dataset including first units and the second dataset including second units, includes indicating a distributing probability of each of the first units onto each of the second units in one distributing matrix of the pair of distributing matrices and indicating a distributing probability of each of the second units onto each of the first units in another distributing matrix of the pair of distributing matrices.

5. The method according to claim 1 , further comprising developing the spatial relationships within each of the multi-scale spatial datasets as binary or continuous matrices.

6. The method according to claim 4 , wherein the developing the spatial relationship among each pair of units within a dataset of the multi-scale spatial datasets includes indicating, in the associated binary matrix, at least one of whether the pair of units are neighbors, whether the pair of units is directly connected, whether the pair of units is within a threshold Euclidean distance of each other, or whether the pair of units is within a threshold path distance of each other.

7. The method according to claim 4 , wherein the developing the spatial relationship among each pair of units within a dataset of the multi-scale spatial datasets includes indicating, in the associated continuous matrix, at least one of a Euclidean distance between the pair of units, a path distance between the pair of units, or a distance between the pair of units when a detour around an obstruction must be taken.

8. The method according to claim 1 , wherein the selecting the features of interest includes discarding from selection the new features with a spatial variation outside a predefined range.

9. The method according to claim 8 , further comprising measuring an importance of each of the new features remaining after the discarding, the importance being based on a spatial correlation to a target of the spatio-temporal prediction.

10. A system to perform spatio-temporal prediction, the system comprising:
an input interface configured to receive multi-scale spatial datasets from one or more sources, each of the multi-scale spatial datasets providing a type of information at a corresponding granularity, at least two of the multi-scale spatial datasets providing at least two different types of information at different corresponding granularities; and
a processor configured to generate spatial relationships both between the multi-scale spatial datasets at the different corresponding granularities and within the multi-scale spatial datasets, to generate features from each of the multi-scale spatial datasets, wherein each feature of each of the multi-scale spatial datasets is a unit of the multi-scale spatial dataset, to generate new features for each of the multi-scale spatial datasets, the new features being based on the features of each of the multi-scale spatial datasets and the spatial relationships between and within the multi-scale spatial datasets, to select features of interest from among the new features, to train a predictive model based on the features of interest, and to predict an event based on the predictive model.

11. The system according to claim 10 , wherein at least two of the multi-scale spatial datasets are in at least one pairing.

12. The system according to claim 11 , wherein the processor develops the spatial relationship between each pairing of the multi-scale spatial datasets as a pair of distributing matrices.

13. The system according to claim 12 , wherein the processor develops the spatial relationship between a pairing of a first dataset and a second dataset among the multi-scale spatial datasets, the first dataset including first units and the second dataset including second units, based on indicating a distributing probability of each of the first units onto each of the second units in one distributing matrix of the pair of distributing matrices and indicating a distributing probability of each of the second units onto each of the first units in another distributing matrix of the pair of distributing matrices.

14. The system according to claim 10 , wherein the processor develops the spatial relationship within each of the multi-scale spatial datasets as binary or continuous matrices.

15. The system according to claim 14 , wherein the processor develops the spatial relationship among each pair of units within a dataset of the multi-scale spatial datasets based on indicating, in the associated binary matrix, at least one of whether the pair of units are neighbors, whether the pair of units is directly connected, whether the pair of units is within a threshold Euclidean distance of each other, or whether the pair of units is within a threshold path distance of each other.

16. The system according to claim 14 , wherein the processor develops the spatial relationship among each pair of units within a dataset of the multi-scale spatial datasets based on indicating, in the associated continuous matrix, at least one of a Euclidean distance between the pair of units, a path distance between the pair of units, or a distance between the pair of units when a detour around an obstruction must be taken.

17. The system according to claim 10 , wherein the processor selects the features of interest based on discarding from selection the new features with a spatial variation outside a predefined range.

18. The system according to claim 17 , further comprising measuring an importance of each of the new features remaining after the discarding, the importance being based on a spatial correlation to a target of the spatio-temporal prediction.

19. A non-transitory computer program medium comprising instructions that, wh

Assignees

• IBM

Inventors

• Dong Wei Shan
• Hampapur Arun
• Li Hongfei
• Li Li
• Liu Xuan
• Ma Chun Yang
• Xing Songhua

Classifications

• G06F16/2477

  Temporal data queries · CPC title

• G06F16/9537

  Spatial or temporal dependent retrieval, e.g. spatiotemporal queries · CPC title

• G06F16/2237

  Vectors, bitmaps or matrices · CPC title

• G06N7/005

  Physics · mapped topic

• G06N99/005 Primary

  Physics · mapped topic