Multi-scale multi-granularity spatial-temporal traffic volume prediction
Methods and systems for allocating network resources responsive to network traffic include modeling spatial correlations between fine spatial granularity traffic and coarse spatial granularity traffic for different sites and regions to determine spatial feature vectors for one or more sites in a network. Temporal correlations at a fine spatial granularity are modeled across multiple temporal scales, based on the spatial feature vectors. Temporal correlations at a coarse spatial granularity are modeled across multiple temporal scales, based on the spatial feature vectors. A traffic flow prediction is determined for the one or more sites in the network, based on the temporal correlations at the fine spatial granularity and the temporal correlations at the coarse spatial granularity. Network resources are provisioned at the one or more sites in accordance with the traffic flow prediction.
1. A method for allocating network resources responsive to network traffic, comprising:
modeling spatial correlations between fine spatial granularity traffic and coarse spatial granularity traffic for a plurality of different sites and regions to determine spatial feature vectors for one or more sites in a network, using a processor-implemented neural network;
modeling temporal correlations at a fine spatial granularity across multiple temporal scales, based on the spatial feature vectors;
modeling temporal correlations at a coarse spatial granularity across multiple temporal scales, based on the spatial feature vectors;
determining a traffic flow prediction for the one or more sites in the network, based on the temporal correlations at the fine spatial granularity and the temporal correlations at the coarse spatial granularity; and
provisioning network resources at the one or more sites in accordance with the traffic flow prediction.
2. The method of claim 1 , wherein modeling the spatial correlations includes determining historical traffic flow information using a site-level graph and a region-level grid.
3. The method of claim 2 , wherein each site in the site-level graph represents a particular device and each region in the region-level grid reflects multiple devices.
4. The method of claim 1 , wherein modeling the spatial correlations includes:
determining region-level spatial correlations; and
determining site-level spatial correlations based on the region-level spatial correlations.
5. The method of claim 1 , wherein modeling spatial correlations includes outputting fine spatial granularity spatial feature vectors and coarse spatial granularity spatial feature vectors.
6. The method of claim 5 , wherein modeling temporal correlations at the fine spatial granularity uses the fine spatial feature vectors and modeling temporal correlations at the coarse spatial granularity uses the coarse spatial feature vectors.
7. The method of claim 1 , wherein modeling temporal correlations at each of the fine spatial granularity and the coarse spatial granularity comprises encoding the spatial feature vectors at multiple different temporal scales.
8. The method of claim 7 , wherein encoding the spatial vectors at multiple different temporal scales includes encoding the spatial vectors at scales that include one week, one day, and one hour.
9. The method of claim 1 , further comprising combining the temporal correlations at the fine temporal granularity and the temporal correlations at the coarse temporal granularity, wherein determining the traffic flow prediction is based on the combined temporal correlations.
10. A method for allocating resources responsive to site-level traffic, comprising:
modeling spatial correlations between fine spatial granularity traffic and coarse spatial granularity traffic for a plurality of different sites and regions to determine spatial feature vectors for one or more sites, using a processor-implemented neural network;
modeling temporal correlations at a fine spatial granularity across multiple temporal scales, based on the spatial feature vectors;
modeling temporal correlations at a coarse spatial granularity across multiple temporal scales, based on the spatial feature vectors;
determining a traffic flow prediction for the one or more sites in the network, based on the temporal correlations at the fine spatial granularity and the temporal correlations at the coarse spatial granularity; and
providing resources at the one or more sites in accordance with the traffic flow prediction.