Predictive Model Accuracy via Filter Combination Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing predictive modeling techniques face challenges in improving accuracy and efficiently handling large volumes of data, requiring significant computational resources and expertise, especially when selecting and tuning effective models.
Innovation Solution
A computer-implemented method that involves generating multiple modified training data sets using different filter combinations, training predictive models, determining their accuracy, and identifying the most accurate model, which can then be applied to new data sets with similar characteristics, along with the use of a predictive modeling server system to scale processes across multiple computers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple filter combinations are applied to generate modified training data sets, then predictive model accuracy is improved, but computational resources and processing time increase
Solution Approach 1:
The patent applies filters to training data in advance to generate multiple modified training data sets before model training. This preliminary action allows the system to explore different data representations and identify optimal filter combinations that improve predictive accuracy without requiring extensive computational resources during the actual prediction phase.
Solution Approach 2:
The system applies multiple filters beyond what a single filter would provide, generating multiple modified training data sets. This excessive action of applying more filters than minimally required allows the system to find the optimal combination that maximizes predictive accuracy, with the benefit that only the best combination needs to be used in production.
2Measurement precision
If multiple predictive models are trained with different filter combinations, then model accuracy improves, but processing time increases
Solution Approach 1:
The patent performs filter combination exploration and model training in advance during an offline phase. By completing the computationally intensive task of training multiple models with different filter combinations beforehand, the system establishes optimal filtering strategies that can be quickly applied to new data without requiring extensive processing time during online operations.
Solution Approach 2:
The patent divides the data processing workflow into distinct phases: an offline training phase where multiple filter combinations are explored and models are trained, and an online prediction phase where the optimized model is deployed. This segmentation allows computationally intensive operations to be performed when resources are available, while maintaining fast response times during actual prediction operations.
3Measurement precision
If filter combinations are optimized for each training data set, then predictive accuracy improves, but system complexity increases
Solution Approach 1:
The patent implements an automated system that independently explores different filter combinations, trains multiple models, evaluates their performance, and identifies the optimal filter combination without requiring manual intervention. This self-service approach handles the complexity of optimizing filter combinations for different data sets automatically, reducing the burden on users while maintaining high predictive accuracy.
Solution Approach 2:
The patent creates a universal framework that can handle different types of training data and automatically determine appropriate filter combinations. The system is designed to work with various data types (text, numerical, etc.) and applies different filtering strategies (n-gram, stopword, stemming, etc.) based on the data characteristics, making the complex optimization process accessible and effective across multiple应用场景.
4Ease of operation
If cloud-based resources are used for training and model selection, then computational burden on client systems is reduced, but network dependency increases
Solution Approach 1:
The patent introduces a cloud-based predictive modeling server as an intermediary between the client system and the complex model training process. The server handles the computationally intensive tasks of applying filter combinations, training multiple models, and selecting the optimal model, while the client system simply needs to provide data and receive predictions. This intermediary approach reduces the computational burden on client systems while maintaining ease of operation.
Data Source
AI summary
In general, a method includes receiving a training data set that includes a plurality of examples, wherein each example includes one or more features and an answer, generating a plurality of modified training data sets by applying one or more filters to the training data set, each of the plurality of modified training data sets being based on a different combination of the one or more filters, training a plurality of predictive models, each of the plurality of predictive models being trained using a different modified training data set of the plurality of modified training data sets, determining a respective accuracy for each of the plurality of predictive models, identifying a most accurate predictive model based on the determined accuracies, and specifying an association between the training data set and the combination of filters used to generate the modified training data set that was used to train the most accurate predictive model.