Time Series Prediction Visualization Using Additive Decomposition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Time series prediction models often fail to provide clear and intuitive visualizations of the factors contributing to predicted future values, making it difficult for users to understand and interpret the predictions, especially when using differencing techniques that model variations rather than the original time series data.
Innovation Solution
The solution involves rewriting the prediction equation to represent it as an equivalent additive expression that separates the trend and cyclic components, ensuring these components are in the same value space as the original time series, allowing for a clear visualization of how predicted values are calculated by summing trend and cyclic components.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If differencing techniques are used to model variations in time series data, then prediction accuracy is improved, but visualization clarity and user interpretability deteriorate
Solution Approach 1:
The patent introduces an intermediary representation that bridges the differenced model and original time series data. By adding the differenced values back to the original baseline, the system creates a visual mapping that connects the modeled variations to the actual time series context, allowing users to interpret predictions in terms of the original data scale and patterns
Solution Approach 2:
The patent transforms the visualization by introducing a new dimensional perspective that displays both the differenced components and their cumulative effect on the original time series. This multi-dimensional approach allows users to see both the detailed variations captured by differencing and the overall trend in the original data context simultaneously
2Reliability
If differencing techniques are used to model variations, then the model captures data patterns more effectively, but the visualization becomes less intuitive and harder to interpret
Solution Approach 1:
The patent segments the visualization into distinct components: the original baseline time series, the differenced variations, and the cumulative reconstructed values. This segmentation allows each component to be displayed and interpreted separately while showing their relationships, reducing the perceived complexity by breaking down the integrated model into understandable parts
Solution Approach 2:
The patent creates visual copies of the time series data at different levels of transformation. By displaying the original data alongside the differenced and reconstructed versions, users can directly compare and understand the relationship between the modeled variations and the actual data patterns, making the complex model more interpretable
Data Source
AI summary
A system and method including determining, for a calculation of a predicted value specified by a recursive equation, an equivalent representation based on a cumulative sum; determining a trend component for the determined equivalent representation of the predicted value; decomposing the determined equivalent representation of the predicted value into the trend component and a cyclic component; and generating a visualization for a set of time series data expressed by the decomposition.