Collective Matrix Factorization for Diagnostic Engine Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional diagnostic engines in call centers are costly to maintain, suffer from low recall and high precision issues, and are unable to autonomously adapt to new contexts, while stepwise machine learning approaches fail to leverage interdependencies between symptom identification, root cause detection, and solution finding.
Innovation Solution
A diagnostic engine that constructs a collective matrix with dimensions for symptoms, root causes, and solutions, performing collective matrix factorization to embed diagnostic sessions and state descriptors, allowing for holistic diagnosis by leveraging interdependencies between these elements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a rules-based diagnostic engine is used, then diagnostic accuracy can be maintained, but the system becomes costly to design and maintain as rules need to be manually generated and kept up-to-date
Solution Approach 1:
The patent replaces the mechanical rules-based system with a machine learning model that automatically learns diagnostic patterns from data. Instead of manually creating and maintaining rules, the system uses training data to generate a model that performs diagnostic reasoning, substituting manual rule maintenance with automated learning.
Solution Approach 2:
The machine learning model autonomously learns and adapts to new diagnostic patterns without requiring manual rule updates. The system self-improves by training on new data, automatically adjusting to changes in product lines and diagnostic scenarios without human intervention in rule creation.
2Adaptability or versatility
If Bayesian or heuristic algorithms are used, then the system is less rigid than rule-based approaches, but it still follows a rigid sequence of symptom identification, root cause identification, and solution proposal
Solution Approach 1:
The patent implements a dynamic diagnostic process where the machine learning model can adapt its reasoning sequence based on the specific case. Instead of following a fixed sequence of symptom identification, root cause identification, and solution proposal, the model dynamically determines the optimal diagnostic path based on learned patterns from training data.
Solution Approach 2:
The patent adds a new dimension to the diagnostic process by incorporating multiple possible root causes and solutions simultaneously in the factorization model. Instead of sequentially determining one root cause at a time, the system considers multiple hypotheses in parallel and uses matrix factorization to identify the most likely combination of symptoms, root causes, and solutions.
3Productivity
If highly skilled experts staff the call center, then performance is maximized, but costs increase and there may not be enough experts to adequately staff the center
Solution Approach 1:
The patent introduces an automated diagnostic support system as an intermediary between call center staff and customers. This system acts as a knowledge assistant that provides real-time diagnostic guidance to less experienced staff, enabling them to perform at expert levels without requiring actual expert staffing.
Solution Approach 2:
The machine learning model captures and replicates expert diagnostic knowledge in a computational form. By training on expert diagnostic data, the system creates a digital copy of expert reasoning capabilities that can be deployed at scale without incurring the costs of hiring and training actual human experts.
4Device complexity
If less well trained staffers staff the call center, then costs are reduced, but performance decreases
Solution Approach 1:
The automated diagnostic support system serves as a knowledge intermediary that bridges the gap between less experienced staff and expert-level performance. The system provides real-time guidance and decision support, enabling less trained staffers to deliver high-quality diagnostic services without requiring extensive training programs.
Data Source
AI summary
A collective matrix is constructed, having a diagnostic sessions dimension and a diagnostic state descriptors dimension. The diagnostic state descriptors dimension includes a plurality of symptom fields, a plurality of root cause fields, and a plurality of solution fields. Collective matrix factorization of the collective matrix is performed to generate a factored collective matrix comprising a sessions factor matrix embedding diagnostic sessions and a descriptors factor matrix embedding diagnostic state descriptors. An in-progress diagnostic session is embedded in the factored collective matrix. A symptom or solution is recommended for evaluation in the in-progress diagnostic session based on the embedding. The diagnostic state descriptors dimension may further include at least one information field storing a representation (for example, a bag-of-words representation) of a semantic description of a problem being diagnosed by the in-progress diagnostic session.