Cognitive Forecasting System Multivariate Correlation Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current forecasting methods are inadequate in accounting for correlations with influential time-varying factors, leading to inaccurate forecasts and suboptimal resource allocation, resulting in unnecessary expenses and poor user experiences across various industries.
Innovation Solution
A cognitive forecasting system that organizes input data into multidimensional attribute spaces, generates time-series forecasts, runs diagnostic tests, calculates correlations between variables, and adjusts forecasts using neural networks to account for correlations between different input bands, thereby improving forecast accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional forecasting models (exponential smoothing or ARIMA) are used, then the forecasting process is relatively simple and straightforward, but the forecast accuracy is poor because these models do not account for correlations with influential time-varying factors
Solution Approach 1:
The patent transforms the forecasting approach from univariate time series analysis to multivariate analysis by organizing input data into multiple bands representing different dimensions (e.g., different time scales, variable types, or feature categories). This dimensional expansion allows the model to capture correlations across multiple factors simultaneously, thereby improving forecast accuracy while managing complexity through structured organization of the additional dimensions.
Solution Approach 2:
The patent implements a nested forecasting architecture where multiple forecasting models operate at different levels or scales. The system generates forecasts at various time horizons and scales them to a common reference point, with each level nested within the broader forecasting framework. This nested structure allows complex correlations to be captured at different levels while maintaining an organized, manageable system architecture.
2Measurement precision
If more complex models with multiple variables are used to account for correlations, then forecast accuracy improves, but the computational resources and time required increase significantly
Solution Approach 1:
The patent performs preliminary organization and preprocessing of input data into structured bands before the actual forecasting computation. By pre-organizing data into meaningful dimensions and scales, the system reduces the computational burden during the forecasting phase. This preliminary structuring allows complex multivariate correlations to be processed more efficiently, reducing the time required for the actual forecasting computation while maintaining accuracy.
3Ease of manufacture
If traditional forecasting models are used, then the implementation is straightforward and inexpensive, but resource allocation is suboptimal leading to unnecessary expenses
Solution Approach 1:
The patent creates a universal forecasting framework that can handle multiple types of input data across different dimensions and scales through a single integrated system. The multi-scale forecasting capability and correlation analysis mechanisms can be applied to various forecasting scenarios (different industries, different variable types, different time horizons) without requiring separate specialized models, thereby improving resource allocation efficiency across diverse applications while maintaining ease of implementation through a standardized approach.
Data Source
AI summary
In an embodiment, a correlated forecast is computer generated by a processor that receives input data for historic values of a first input variable, creates forecast data for future values of the first input variable using the historic values of the first input variable, generates diagnostic data based on a diagnostic analysis of the forecast data, creates a first diagnostic variable that includes a first diagnostic value from a first cognitive process, generates a feature vector based on a second cognitive process that determines the feature vector by identifying a correlation between the first diagnostic variable and a second diagnostic variable, and generates a final forecast using the feature vector as an input for a cognitive forecasting process, where the first cognitive process determines the first diagnostic value based on the diagnostic data.