Unified ML Model for Record Topic Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning algorithms for predicting record topics face challenges due to limited and low-quality data, inconsistent terminology, and variable formats in transaction records, leading to degraded performance and the need for extensive human intervention and multiple models for each user account.
Innovation Solution
A computer-implemented method generates a dataset using transitive relations among historical records, forming pairs of antecedent and consequent topics to predict next record topics, using a single machine learning model across multiple accounts, which improves data efficiency and mitigates the cold start problem.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple individual ML models are created for each user account to handle data variability, then prediction accuracy for each user may improve, but system complexity and resource requirements increase significantly
Solution Approach 1:
The patent combines multiple user-specific models into a single unified ML model that processes all user data through one system. This consolidation reduces system complexity while maintaining prediction accuracy by learning from aggregated patterns across all users rather than requiring separate models for each account.
Solution Approach 2:
The unified ML model serves multiple functions by handling prediction tasks for all users simultaneously. It processes diverse data formats and topics from different users through a single versatile system, eliminating the need for user-specific model instances while maintaining adaptability to individual user patterns.
2Reliability
If extensive data preprocessing and human intervention are applied to handle inconsistent terminology and formats, then data quality improves, but processing time and operational complexity increase
Solution Approach 1:
The patent transforms the data processing approach by changing parameters from manual preprocessing to automated ML-based processing. The system adapts to inconsistent terminology and formats dynamically during training, learning to map varied inputs to meaningful predictions without requiring extensive manual data cleaning or standardization steps.
Solution Approach 2:
The ML model performs self-service by automatically handling data variability and inconsistency during training. It learns to process diverse data formats, abbreviations, and terminology without requiring external human intervention for data preprocessing, thereby reducing operational complexity and processing time.
3Productivity
If a single ML model is used across multiple accounts to reduce complexity, then system efficiency improves, but the ability to capture user-specific patterns may deteriorate
Solution Approach 1:
The patent segments the training data by user account while maintaining a unified model architecture. The model learns from segmented user-specific patterns during training but consolidates these learnings into a single system that can efficiently process any user's data, capturing individual patterns without requiring separate model instances.
Solution Approach 2:
The patent adds a user identity dimension to the model's learning process. By incorporating user-specific identifiers and patterns as additional features during training, the unified model captures individual user behaviors and preferences while maintaining system-wide efficiency through a single consolidated architecture.
Data Source
AI summary
A method includes generating dataset using topics associated with historical records, the dataset including pairs of data that are formed based on the topics, each of the pairs of data including an antecedent topic associated with a historical record corresponding to a preceding event and a consequent topic associated with a historical record corresponding to an event that occurred after the preceding event, the antecedent topic and the consequent topic forming a transitive relation for each of the pairs of data; inputting, into ML model, the pairs of data and input topic associated with a record of a user; generating, by the ML model, a prediction of a next record topic for a next record corresponding to the user, based on the consequent topic included in each of the pairs of data that include the antecedent topic corresponding to the input topic; and outputting the prediction.