Attention-Based Neural Network Forecasting Causal Impact
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current predictive analytics models, particularly neural networks, lack interpretability and fail to effectively quantify the relative causal impact of external factors on time-series data, limiting their adoption in industries.
Innovation Solution
A computing device and method utilizing attention-based neural networks to forecast future behavioral data and identify the relative causal impact of external factors by creating statistical prediction models that incorporate external data, providing an interpretable additive effect on baseline forecasts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If neural networks are used for predictive modeling, then predictive accuracy is improved, but interpretability deteriorates
Solution Approach 1:
The patent segments the predictive modeling process into two distinct components: a neural network component for capturing complex patterns and relationships in the data, and a statistical model component for providing interpretable results. This segmentation allows each component to specialize - the neural network handles accuracy while the statistical model handles interpretability - thereby resolving the contradiction between these two competing requirements.
Solution Approach 2:
The patent introduces an intermediary statistical model that acts as a bridge between the neural network's complex internal representations and the need for human-interpretable results. The statistical model translates the neural network's predictions into forms that can be explained in terms of causal relationships and relative impacts, thus mediating between the accuracy provided by neural networks and the interpretability required by users.
2Loss of information
If traditional statistical models are used, then interpretability is improved, but predictive accuracy deteriorates
Solution Approach 1:
The patent merges two previously separate approaches - neural networks and statistical models - into a unified hybrid framework. The neural network processes the input data and generates predictions, while the statistical model simultaneously processes the same data to generate interpretable explanations. By merging these two approaches, the system achieves both high predictive accuracy (from the neural network) and strong interpretability (from the statistical model), resolving the contradiction between these two features.
3Measurement precision
If multiple external factors are incorporated into the model, then predictive accuracy is improved, but device complexity deteriorates
Solution Approach 1:
The patent extracts and isolates the complexity of handling multiple external factors into a dedicated statistical model component. Rather than having the neural network directly process and interpret all external factors, the statistical model takes out the task of analyzing external factor impacts and presents them in a simplified, interpretable format. This extraction reduces the overall system complexity while maintaining the ability to incorporate multiple external factors for improved predictive accuracy.
Data Source
AI summary
A method and system are provided to calculate a future behavioral data and identify a relative causal impact of external factors affecting the data. Behavioral data and data for one or more external factors are harvested for a first time period. New behavioral data is harvested for a second time period. New data for the second time period is harvested. Based on a second training algorithm, a forecast time series value of a future behavioral data for a third time period that is after the second time period is calculated. A relative causal impact between each external factor and the predicted time series value of the behavioral data, for the third time period, is identified.