Hierarchical Latent Variable Model Estimation for Supply Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for predicting supply amounts, such as those described in Japanese Patent No. 4139410 and Japanese Unexamined Patent Application 2010-128779, face challenges in model classification and reliability due to the lack of consideration for hierarchical latent variables, leading to inadequate computational procedures and theoretical justification.
Innovation Solution
A hierarchical latent variable model estimation device and method that inputs data for study, sets a hierarchical latent structure with tree representation, computes variational probabilities, optimizes components, and optimizes gating functions to determine branch directions, enabling accurate supply amount predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If prediction models are classified based on expert opinion for different days, seasons, and weather information, then the system can provide customized predictions, but it becomes difficult to set an appropriate reference and reduces reliability
Solution Approach 1:
The system automatically determines the optimal prediction model by having the data itself indicate which model should be used, rather than relying on expert opinion. The determination unit selects from multiple prediction models based on learned patterns in the data, enabling the system to self-optimize without external guidance while maintaining high reliability
Solution Approach 2:
The system changes the parameter of model selection from static expert-defined categories to dynamic data-driven selection. By learning from historical data which models perform best under various conditions, the system adapts the model selection parameter automatically, achieving both customization and reliability
2Ease of manufacture
If the method from Factorized Asymptotic Bayesian Inference is used for mixture modeling, then model estimation can be performed, but it cannot solve model selection problems for hierarchical latent variables and loses theoretical justification
Solution Approach 1:
The system introduces a determination unit as an intermediary between the multiple prediction models and the final prediction output. This intermediary component learns the optimal selection strategy from data and acts as a bridge that connects different models to appropriate usage scenarios, enabling both estimation capability and theoretical soundness
Solution Approach 2:
The system segments the prediction task into multiple independent prediction models, each optimized for specific conditions. Rather than using a single monolithic model, the system divides the problem into manageable segments and uses the determination unit to select the appropriate segment based on current conditions, solving both estimation and selection problems
Data Source
AI summary
A hierarchical latent structure setting unit 81 sets a hierarchical latent structure that is a structure in which latent variables are represented by a tree structure and components representing probability models are located at nodes of a lowest level of the tree structure. A variational probability computation unit 82 computes a variational probability of a path latent variable that is a latent variable included in a path linking a root node to a target node in the hierarchical latent structure. A component optimization unit 83 optimizes each of the components for the computed variational probability. A gating function optimization unit 84 optimizes a gating function model that is a model for determining a branch direction according to the multivariate data in a node of the hierarchical latent structure, based on the variational probability of the latent variable in the node.