Computing predicted data according to weighted peak preservation and time distance biasing
A value corresponding to an adjustable control element is received. Predicted data is computed from seasonal data, where the computing is according to applying preservation of peaks in the seasonal data and applying time distance biasing in which more recent data points in the seasonal data are weighted higher than less recent data points. Relative weighting of the peak preservation and the time distance biasing is based on the received value corresponding to the adjustable control element.
1. A method comprising:
receiving seasonal data that exhibits a repeating pattern over time;
receiving, by a system having a processor, a value corresponding to an adjustable control element;
applying, in the seasonal data by the system, preservation of peaks and time distance biasing in which more recent data points in the seasonal data are weighted higher than less recent data points;
calculating, by the system based on the received value corresponding to the adjustable control element, a first weight for the preservation of peaks and a second weight for the time distance biasing;
combining, by the system, the first and second weights to produce a combined weight; and
computing, by the system, predicted data using the combined weight.
2. The method of claim 1 , wherein receiving the value corresponding to the adjustable control element comprises receiving the value corresponding to a user-moveable control icon in a graphical user interface.
3. The method of claim 1 , wherein the computing further comprises:
aggregating the combined weight with data points in the seasonal data to produce the predicted data.
4. The method of claim 1 , wherein calculating the first weight is based on a product of a value derived from the received value and a first weight coefficient for a respective data point in the seasonal data, wherein the first weight coefficient is determined by recursively partitioning the seasonal data according to peaks and using a recursion depth to estimate an importance of a corresponding peak in the seasonal data.
5. The method of claim 4 , wherein calculating the second weight is based on a product of a value derived from the received value and a second weight coefficient for a respective data point in the seasonal data, the second weight coefficient decreasing with increasing time distance from a time point at which data prediction is to be performed.
6. The method of claim 1 , further comprising visualizing a time series in the seasonal data using a cell-based visualization that includes cells representing respective data points in the seasonal data and respective predicted data points in the predicted data, wherein the cells are assigned respective colors according to attribute values of the corresponding data points.
7. The method of claim 1 , further comprising:
displaying a representation of predicted data points in the predicted data; and
displaying a certainty band associated with the predicted data points for indicating degrees of certainty associated with the corresponding predicted data points.
8. The method of claim 7 , wherein the certainty band uses different shading to highlight corresponding ones of the predicted data points that are associated with a more pronounced peak.
9. The method of claim 1 , wherein a first adjustment of the adjustable control element causes the first weight to increase and the second weight to decrease, and wherein a second, different adjustment of the adjustable control element causes the first weight to decrease and the second weight to increase.
10. An article comprising at least one non-transitory machine-readable storage medium storing instructions that upon execution cause a system to:
receive seasonal data that exhibits a repeating pattern over time;
receive an indication relating to adjustment of an adjustable control element;
assign weights based on the indication, wherein the weights include a first weight for peak preservation of peaks in the seasonal data, and a second weight for time distance biasing in which more recent data points in the seasonal data are weighted higher than less recent data points, wherein assigning the weights comprises
calculating, based on a value derived from the received indication, the first weight for the peak preservation of peaks and the second weight for the time distance biasing;
combine the first weight and the second weight to produce a combined weight; and
compute predicted data using the combined weight.
11. The article of claim 10 , wherein the adjustment control element includes a weighting slider.
12. The article of claim 10 , wherein the instructions upon execution cause the system to further:
present a cell-based visualization of the seasonal data; and
receive, based on the cell-based visualization, a selection of a pattern for which the computing of the predicted data is to be performed.
13. The article of claim 12 , wherein the cell-based visualization includes cells representing respective data points of the seasonal data, the cells assigned colors corresponding to a coloring attribute of the data points.
14. The article of claim 10 , wherein the instructions upon execution cause the system to further:
present a line graph representing the seasonal data and the predicted data.
15. The article of claim 14 , wherein the instructions upon execution cause the system to further:
present a certainty band around a portion of the line graph representing the predicted data, where the certainty band indicates degrees of certainty associated with the corresponding predicted data points in the predicted data.
16. The article of claim 10 , wherein the instructions upon execution cause the system to further:
aggregate the combined weight with data points in the seasonal data to produce the predicted data.
17. The article of claim 10 , wherein a first adjustment of the adjustable control element causes the first weight to increase and the second weight to decrease, and wherein a second, different adjustment of the adjustable control element causes the first weight to decrease and the second weight to increase.
18. The article of claim 10 , wherein calculating the first weight is based on the product of a value derived from the received indication and a weight coefficient for a respective data point in the seasonal data, wherein the weight coefficient is determined by recursively partitioning the seasonal data according to peaks and using a recursion depth to estimate an importance of a corresponding peak in the seasonal data.
19. A system comprising:
at least one processor to:
receive seasonal data that exhibits a repeating pattern over time;
receive a value corresponding to an adjustable control element;
calculate, according to the received value, a first weight for preservation of peaks in the seasonal data and a second weight for time distance biasing in which more recent data points in the seasonal data are weighted higher than less recent data points, wherein applying the preservation of peaks avoids or reduces removal of peaks in the seasonal data;
combine the first and second weights to produce a combined weight; and
compute predicted data using the combined weight.
20. The system of claim 19 , wherein the computing is further based on:
aggregating the combined weight with data points in the seasonal data to produce the predicted data.
21. The system of claim 19 , wherein a first adjustment of the adjustable control element causes the first weight to increase and the second weight to decrease, and wherein a second, different adjustment of the adjustable control element causes the first weight to decrease and the second weight to increase.
22. The system of claim 19 , wherein calculating the first weight is based on a product of a value derived from the received value and a weight coefficient for a respective data point in the seasonal data, wherein the weight coefficient is determined by recursively partitioning the seasonal data according to peaks and using a recursion depth to estimate an importance of a corresponding peak in the seasonal data.