Performing multivariate time series prediction with three-dimensional transformations

1. A computer-implemented method, comprising:

receiving historical two-dimensional (2D) multivariate time series data;

transforming the historical 2D multivariate time series data into a three-dimensional (3D) temporal tensor, wherein the 3D temporal tensor includes a time-based geometric object represented by an array of components that are functions of coordinates of a space;

training one or more deep volumetric 3D convolutional neural networks (CNNs), utilizing the 3D temporal tensor; and

predicting future values for additional multivariate time series data, utilizing the one or more trained deep volumetric 3D CNNs.

2. The computer-implemented method of claim 1 , wherein the historical 2D multivariate time series data include 2D multivariate time series data that has been obtained from sensors and recorded in a hardware memory prior to a current time and date.

3. The computer-implemented method of claim 1 , wherein a first dimension of the historical 2D multivariate time series data includes a time dimension, and a second dimension of the historical 2D multivariate time series data includes a feature dimension.

4. The computer-implemented method of claim 1 , comprising performing stack-wise average pooling on output of 3D convolutions from the one or more trained deep volumetric 3D CNNs.

5. The computer-implemented method of claim 1 , comprising applying adaptive hyperparameter tuning to the 3D temporal tensor to determine optimal parameters required for a temporal tensor transformation via a propagation of prediction error.

6. The computer-implemented method of claim 1 , wherein the one or more deep volumetric 3D CNNs each include a residual network-based module used for extraction of non-linear feature-temporal and auto-correlational features.

7. The computer-implemented method of claim 1 , wherein the one or more deep volumetric 3D CNNs discover local temporal dependencies and auto-correlational patterns among values of dimensions of the 3D temporal tensor.

8. The computer-implemented method of claim 1 , wherein during training, volumetric convolutions of the one or more deep volumetric 3D CNNs learn a three-dimensional kernel.

9. The computer-implemented method of claim 1 , wherein the additional multivariate time series data include incomplete 2D multivariate time series data having a time dimension, the incomplete 2D multivariate time series data also having a missing or incomplete feature.

10. The computer-implemented method of claim 1 , wherein the one or more trained deep volumetric 3D CNNs utilize identified local temporal dependencies and auto-correlational patterns among values of dimensions learned during training to determine one or more feature maps for the additional multivariate time series data.

11. The computer-implemented method of claim 1 , comprising using the predicted values to allocate bandwidth and processor resources to an application running on a device thereby improving performance of the device and the application.

12. The computer-implemented method of claim 1 , wherein the transforming includes dividing the historical 2D multivariate time series data into a plurality of 2D temporal slices, wherein the transforming includes stacking the plurality of 2D temporal slices to create the 3D temporal tensor.

13. The computer-implemented method of claim 1 , and comprising utilizing the predicted future values to discover a new trend, to detect anomalies, to predict future signals, to discover a new trend and detect anomalies, to discover a new trend and predict future signals, to detect anomalies and predict future signals, or to detect anomalies, discover a new trend and predict future signals.

14. A computer program product for performing multivariate time series prediction with three-dimensional transformations, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to:

receive, utilizing the processor, historical two-dimensional (2D) multivariate time series data;

transform, utilizing the processor, the historical 2D multivariate time series data into a three-dimensional (3D) temporal tensor, wherein the transforming includes dividing the historical 2D multivariate time series data into a plurality of 2D temporal slices, wherein the transforming includes stacking the plurality of 2D temporal slices to create the 3D temporal tensor;

train, utilizing the processor, one or more deep volumetric 3D convolutional neural networks (CNNs), utilizing the 3D temporal tensor; and

predict future values for additional multivariate time series data, utilizing the processor and the one or more trained deep volumetric 3D CNNs.

15. The computer program product of claim 14 , wherein the historical 2D multivariate time series data include 2D multivariate time series data that has been recorded prior to a current time and date.

16. The computer program product of claim 14 , wherein a first dimension of the historical 2D multivariate time series data includes a time dimension, and a second dimension of the historical 2D multivariate time series data includes a feature dimension.

17. The computer program product of claim 14 , comprising program instructions executable by a processor to cause the processor to perform stack-wise average pooling on output of 3D convolutions from the one or more trained deep volumetric 3D CNNs.

18. The computer program product of claim 14 , comprising program instructions executable by a processor to cause the processor to apply adaptive hyperparameter tuning to the 3D temporal tensor to determine optimal parameters required for a temporal tensor transformation via a propagation of prediction error.

19. A system, comprising:

a hardware processor;

hardware memory coupled to the processor; and

logic integrated with the processor, executable by the processor, or integrated with and executable by the processor, the logic being configured to:

receive historical two-dimensional (2D) multivariate time series data;

transform the historical 2D multivariate time series data into a three-dimensional (3D) temporal tensor, wherein the 3D temporal tensor includes a time-based geometric object represented by an array of components that are functions of coordinates of a space;

train one or more deep volumetric 3D convolutional neural networks (CNNs), utilizing the 3D temporal tensor; and

predict future values for additional multivariate time series data, utilizing the one or more trained deep volumetric 3D CNNs.