Hierarchical Data Prediction System With Feedback-Driven Model Retraining
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional enterprise software and machine learning systems face limitations such as human error, inadequate data, personal biases, and computation complexity, which affect accuracy and efficiency in data prediction and organizational operations.
Innovation Solution
A computerized method and system for data prediction using a hierarchical data structure with machine learning models that allow for user feedback-driven updates of configuration parameters, enabling re-training of models to improve prediction accuracy and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning models are used for data prediction, then prediction accuracy can be improved, but computation complexity increases
Solution Approach 1:
The patent segments the monolithic machine learning model into multiple specialized models (e.g., classification model, regression model, clustering model) that can be selected and executed independently based on the specific prediction task. This segmentation reduces the computational burden of maintaining and training a single complex model while preserving prediction accuracy through appropriate model selection.
Solution Approach 2:
The patent dynamically adjusts model parameters such as training data subsets, hyperparameters, and model architecture configurations based on the specific prediction requirements and available computational resources. This allows the system to optimize the balance between prediction accuracy and computation complexity by selecting appropriate parameter settings for different scenarios.
2Reliability
If more training data is collected to improve prediction accuracy, then prediction reliability increases, but data processing time and computational resources increase
Solution Approach 1:
The patent implements partial action by training multiple specialized models on different subsets of training data rather than training one comprehensive model on the entire dataset. Each model is trained on a specific portion or type of data relevant to its function, which reduces the time and computational resources required for each individual training process while collectively achieving high prediction reliability through model ensembling or selection.
3Stability of the object's composition
If conventional enterprise software is used, then system stability is maintained, but efficiency and speed of operations decrease
Solution Approach 1:
The patent incorporates feedback mechanisms where prediction results are continuously evaluated against actual outcomes, and model performance metrics are used to automatically adjust and improve future predictions. This feedback loop enables the system to maintain stability through controlled iterative improvements while significantly enhancing operational efficiency by automating decision-making processes and reducing manual intervention.
Data Source
AI summary
There is provided a system and method of data prediction. The method includes obtaining a hierarchical data structure comprising a plurality of layers, each including one or more nodes; obtaining one or more machine learning (ML) models each corresponding to a respective node of at least some of the nodes in at least a given layer, in response to a user's request of prediction related to a given node in the given layer; generating a prediction result using a given ML model corresponding to the given node; upon receiving the user's feedback, selecting one or more configuration parameters of the given ML model related to the feedback; updating the selected configuration parameters according to additional factors in the feedback, and re-training the given ML model to obtain a re-trained ML model,—and using the re-trained ML model to generate an updated prediction result to be sent to the user.