Demand Prediction Modeling With Interpretable Feature Contributions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing demand prediction neural networks lack interpretability, leading to risks in replenishment decisions due to their uninterpretable nature.
Innovation Solution
Implement a method and apparatus that involves determining object categories and generating feature demand prediction models, using interpretable and uninterpretable models to solve the aforementioned technical problems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a pre-trained demand prediction neural network is used to predict demand information, then prediction accuracy can be improved, but the prediction process becomes uninterpretable leading to replenishment risks
Solution Approach 1:
The patent segments the demand prediction process into multiple independent feature prediction models (e.g., price feature model, promotion feature model, seasonal feature model). Each model predicts a specific feature's demand contribution separately, making the prediction process interpretable while maintaining accuracy through aggregation of these segmented predictions.
Solution Approach 2:
The patent introduces feature demand prediction information as an intermediary between the input historical data and the final demand prediction. This intermediary layer provides interpretable feature-level insights that bridge the gap between black-box neural network predictions and understandable business logic.
2Loss of information
If an interpretable demand prediction model is used, then the prediction process becomes transparent, but prediction accuracy may be compromised
Solution Approach 1:
The patent merges multiple interpretable feature prediction models into a unified demand prediction framework. By combining predictions from price feature models, promotion feature models, seasonal feature models, and other interpretable components, the system achieves both interpretability and high prediction accuracy through the aggregation of multiple specialized models.
Solution Approach 2:
The patent creates a composite prediction system that integrates multiple types of feature prediction models (price, promotion, seasonal, and other business-specific features). This composite approach combines the strengths of different interpretable models to achieve overall high accuracy while maintaining transparency through the modular structure.
Data Source
AI summary
A method and apparatus of generating a prediction information, a device, a medium and a program product are provided. The method includes: acquiring feature data corresponding to a target object for a plurality of object demand influence features; determining an object category corresponding to the target object according to the feature data; determining at least one information to be predicted for the target object according to the object category; generating at least one first feature demand prediction information for a target time according to at least one first feature demand prediction model and the feature data; inputting the at least one first feature demand prediction information and the feature data into a pre-trained second feature demand prediction model, so as to generate at least one second feature demand prediction information and a total demand prediction information for the target time.