Predictive Model Selection via Feature Vector Similarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing predictive modeling technologies face challenges in efficiently selecting and applying compatible predictive models to input datasets, particularly when features do not match exactly, leading to difficulties in identifying suitable models for generating predictive outputs.
Innovation Solution
A computer-implemented method and system that utilize a predictive modeling API to identify and select compatible predictive models by matching tags associated with input dataset columns, allowing for the application of predictive models to generate outputs and providing a list of compatible models to client devices, even when features are similar but not identical.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional predictive model selection methods are used, then model compatibility can be ensured, but the process becomes inefficient and time-consuming when features do not match exactly
Solution Approach 1:
The system changes the parameter of feature matching from exact match to similarity-based match using cosine similarity calculation. This allows models to be selected based on feature similarity rather than requiring identical feature names, thereby improving selection efficiency while maintaining compatibility through similarity thresholds
Solution Approach 2:
The system introduces an intermediary mechanism (feature similarity calculation and mapping process) between the input features and model requirements. This intermediary layer translates and matches features based on similarity, enabling efficient model selection even when feature names or structures differ, thus resolving the contradiction between efficiency and compatibility
2Adaptability or versatility
If a large collection of predictive models is maintained, then model versatility is improved, but the complexity of selecting the appropriate model increases
Solution Approach 1:
The system performs preliminary actions by pre-calculating and storing feature vectors for all models in the collection. When a new input arrives, the system quickly computes similarity scores against these pre-prepared vectors, enabling efficient model selection from a large collection without manually reviewing each model's requirements, thus reducing selection complexity while maintaining versatility
Solution Approach 2:
The system transforms the model selection problem from a complex categorical matching task into a simpler numerical similarity comparison task. By representing both input features and model requirements as vectors and using cosine similarity, the system simplifies the selection process from examining multiple model specifications to computing a single similarity score, reducing complexity while handling large model collections
Data Source
AI summary
A computer-implemented method includes obtaining a database table, the database table including data arranged in a plurality of rows and a plurality of columns, each column of data being associated with a different tag that specifies a category for data in the column, using one or more processors to identify a first predictive model, from a collection of predictive models, that can be applied to the database table to generate a predictive output, in which identifying the first predictive model is based on one or more of the different tags, adding a name associated with the first predictive model to a set of names of predictive models that are compatible with the database table, and providing the set of names of predictive models to a client device.