InRule Makes Big Data More Accessible for Automation with Downsampling and Calibration in Machine Learning

InRule Technology, an intelligence automation company providing integrated decisioning, machine learning and process automation software to the enterprise, announced that its suite of machine learning modeling engines now includes integrated downsampling and model calibration, allowing for faster model creation without sacrificing the interpretability of model outputs.

InRule’s downsampling capability reduces the overall size of a data set, allowing teams to train machine learning (ML) models more quickly when working with data sets where the prediction target rarely occurs, while reducing cloud costs for Model Ops infrastructure.

AiThority: How to Get Started with Prompt Engineering in Generative AI Projects

Complementary to integrated downsampling, InRule Machine Learning users can now leverage new model calibration functionality prior to model deployment. Incorporating model calibration allows users to re-align model confidence outputs to the rate of real-world occurrence, improving actionability and interpretability by adjusting the distortion from downsampling data sets.

According to Danny Shayman, InRule Machine Learning product manager, “Many of our customers’ most common use cases involve highly class-imbalanced data sets and this pair of capabilities will significantly reduce their model training times while retaining full interpretability.”

Read: How AI NFTs Are Unlocking the Democratization of the Digital Economy

[To share your insights with us, please write to sghosh@martechseries.com]