Inventory Decision Engine for OOI VOI Split Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing supply chain management systems face computational, storage, and network resource burdens due to inaccurate manual calculations of OEM-owned inventory (OOI) and vendor-owned inventory (VOI) splits, leading to overstock and parts shortages, exacerbated by limited compute, storage, and network resources in distributed computer networks.
Innovation Solution
Implementing intelligent decision management functionalities using a decision engine that classifies inventory items into clusters based on historical OOI/VOI splits, identifies deviations, applies weights, and re-classifies items to optimize OOI/VOI percentages through active learning and generalized additive models, relieving resource burdens.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual calculations of OOI and VOI splits are used, then simplicity of operation is maintained, but measurement precision and reliability deteriorate
Solution Approach 1:
The decision engine automatically performs clustering, deviation detection, weighting, and re-classification of inventory items without manual intervention. The system serves itself by using historical data to generate predictions and continuously improving through active learning, eliminating the need for manual calculations while maintaining high precision.
Solution Approach 2:
The patent replaces manual mechanical calculation methods with an automated computational system using machine learning algorithms (GAMs, k-means clustering, active learning). This substitution transitions from human-operated arithmetic to algorithm-driven automated decision-making, significantly improving precision while the automation handles the complexity.
2Reliability
If accurate OOI/VOI split predictions are implemented, then reliability improves, but compute and storage resources are consumed
Solution Approach 1:
The decision engine segments inventory items into distinct clusters based on their characteristics and historical data patterns. By grouping similar items together, the system processes data more efficiently, reducing overall computational burden while maintaining accurate predictions for each segment. This segmentation allows parallel processing and reduces the complexity of individual item analysis.
Solution Approach 2:
The system performs preliminary clustering and pattern recognition on historical data before making predictions. By pre-processing and organizing data into meaningful clusters in advance, the system reduces the computational effort required during actual prediction operations, thereby improving reliability while managing resource consumption more effectively.
3Measurement precision
If clustering and re-classification operations are performed, then measurement precision improves, but processing time increases
Solution Approach 1:
The decision engine performs clustering and re-classification operations periodically or at scheduled intervals rather than continuously for every data change. This periodic approach allows the system to maintain updated classifications without constant processing, balancing precision requirements with acceptable processing time intervals. The system can operate in batches or trigger-based modes to optimize timing.
Solution Approach 2:
The classification system is dynamic and adaptive, using active learning to continuously improve clustering accuracy based on new data and feedback. The system adjusts cluster assignments and re-classifies items dynamically as patterns emerge, improving precision over time while the learning process becomes more efficient with each iteration, reducing processing time for subsequent operations.
Data Source
AI summary
A decision management technique comprises obtaining, for a set of items, data indicative of a previous percentage division between a first item type and a second item type for each item in the set of items. The technique further comprises classifying each item in the set of items into one of a plurality of clusters, based on the obtained data, wherein each cluster represents a different percentage division range between the first item type and the second item type. The technique further comprises identifying any items in each cluster that deviate from the percentage division range for the cluster, and then applying weights to any identified items based on the deviation from the percentage division range. The technique further comprises re-classifying any identified items to another cluster in the plurality of clusters based on the applied weights.