Method and system for forecasting in sparse data streams via dense data streams
Methods and systems for forecasting in sparse data streams. In an example embodiment, steps or operations can be implemented for mapping a time series data stream to generate forecast features using a neural network, transforming the forecast features into a space with transformed forecast features thereof using metric learning, clustering the transformed forecast features in a cluster, initializing a forecast learning algorithm with a combination of the transformed forecast features in the cluster corresponding to a sparse data stream, and displaying forecasts in a GUI dashboard with information indicative of how the forecasts were achieved, wherein the mapping, the transforming, the clustering, and the initializing together lead to increases in a speed of the forecasting and computer processing thereof.
1. A method for displaying a graphical user interface (GUI) forecasting in sparse data streams via dense data streams, said method comprising:
mapping a time series data stream to generate forecast features using a recurrent neural network comprising an input layer, a recurrent layer, and an output layer;
transforming said forecast features into a space with transformed forecast features using metric learning facilitated by the recurrent neural network;
clustering said transformed forecast features in at least one cluster, wherein the clustering involves weight based initialization to initialize weights of the recurrent neural network for forecasting in a sparse data stream;
initializing a forecast learning algorithm with a combination of said transformed forecast features in said at least one cluster corresponding to the sparse data stream, wherein a relationship between the sparse data stream and a dense data stream is established by the clustering; and
displaying forecasts in a GUI dashboard with information indicative of how said forecasts were achieved, wherein said mapping, said transforming, said clustering, and said initializing together lead to increases in a speed of said forecasting and computer processing of a computer displaying the GUI dashboard through a computer screen associated with a computer.
2. The method of claim 1 wherein sparse in said sparse data stream is defined as a number of occurrences of an event of interest that is sparsely distributed over time.
3. The method of claim 1 wherein:
the forecast feature comprises a hidden unit representation that captures a crime pattern of a city;
the metric learning includes a distance metric satisfying a non-negativity and a triangle inequality; and
said clustering of said transformed forecast features in said at least one cluster is facilitated by a clustering algorithm.
4. The method of claim 3 wherein said clustering algorithm comprises a k-means clustering algorithm.
5. The method of claim 3 wherein said clustering algorithm comprises at least one of: a spectral clustering algorithm, a Mean-Shift Clustering algorithm, a Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm, an Expectation-Maximization (EM) Clustering algorithm using a Gaussian Mixture Model (GMM), and an Agglomerative Hierarchical Clustering algorithm.
6. The method of claim 1 wherein said time series data stream is mapped to fixed length vectors comprising said forecast features, the fixed length vectors including the distance metric.
7. The method of claim 1 wherein a hidden layer is contained with the recurrent layer and the input layer includes embedding blocks and timing and marker inputs.
8. The method of claim 7 wherein said recurrent neural network is based on a (RMTPP) Recurrent Marked Temporal Point Process model, wherein the embedding blocks include an embedded vector, where the embedded vector and temporal features are fed into the recurrent layer.
9. The method of claim 1 wherein said initializing said forecast learning algorithm, further comprises initializing said forecast learning algorithm with a combination of weights from said at least one cluster.
10. A system for displaying a graphical user interface (GUI) forecasting in sparse data streams via dense data streams, said system comprising:
at least one processor;
a memory coupled to said at least one processor;
a set of computer program instructions stored in said memory and executed by said at least one processor in order to perform actions of:
mapping a time series data stream to generate forecast features using a recurrent neural network comprising an input layer, a recurrent layer, and an output layer;
transforming said forecast features into a space with transformed forecast features using metric learning facilitated by the recurrent neural network;
clustering said transformed forecast features in at least one cluster, wherein the clustering involves weight based initialization to initialize weights of the recurrent neural network for forecasting in a sparse data stream;
initializing a forecast learning algorithm with a combination of said transformed forecast features in said at least one cluster corresponding to the sparse data stream, wherein a relationship between the sparse data stream and a dense data stream is established by the clustering; and
displaying forecasts in a GUI dashboard with information indicative of how said forecasts were achieved, wherein said mapping, said transforming, said clustering, and said initializing together lead to increases in a speed of said forecasting and computer processing of a computer displaying the GUI dashboard through a computer screen associated with a computer.
11. The system of claim 10 wherein sparse in said sparse data stream is defined as a number of occurrences of an event of interest that is sparsely distributed over time.
12. The system of claim 10 wherein:
the forecast feature comprises a hidden unit representation that captures a crime pattern of a city;
the metric learning includes a distance metric satisfying a non-negativity and a triangle inequality; and
said clustering of said transformed forecast features in said at least one cluster is facilitated by a clustering algorithm.
13. The system of claim 12 wherein said clustering algorithm comprises a k-means clustering algorithm.
14. The system of claim 12 wherein said clustering algorithm comprises at least one of: a spectral clustering algorithm, a Mean-Shift Clustering algorithm, a Spectral Clustering algorithm, a Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm, an Expectation-Maximization (EM) Clustering algorithm using a Gaussian Mixture Model (GMM), and an Agglomerative Hierarchical Clustering algorithm.
15. The system of claim 10 wherein said time series data stream is mapped to fixed length vectors comprising said forecast features, the fixed length vectors including the distance metric.
16. The system of claim 10 wherein a hidden layer is contained with the recurrent layer and the input layer includes embedding blocks and timing and marker inputs.
17. The system of claim 16 wherein said recurrent neural network is based on a (RMTPP) Recurrent Marked Temporal Point Process model, wherein the embedding blocks include an embedded vector, where the embedded vector and temporal features are fed into the recurrent layer.
18. The system of claim 10 wherein said initializing said forecast learning algorithm, further comprises initializing said forecast learning algorithm with a combination of weights from said at least one cluster.