Browse Prior Art Database

Predictive modeling platform with incremental data updates

IP.com Disclosure Number: IPCOM000243011D
Publication Date: 2015-Sep-08
Document File: 1 page(s) / 184K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is a platform for managing the data, training, and model throughout the software development life cycle. The disclosed platform incorporates individual feedback into the system over time. Such incremental uploads can then trigger another batch of training, potentially of multiple models with multiple parameter sets resulting in the creation of a new version of the model which could then be selected for use in an application.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 62% of the total text.

Page 01 of 1

Predictive modeling platform with incremental data updates
As more applications use trained models and data to produce results rather than static algorithms the need to maintain information about the training data, model parameters, and quality of the trained model results will only increase. Proposed in this disclosure is a platform for managing the data, training, and model throughout the software development life cycle. Similar solutions for such platforms have been documented, however none addresses the need to incorporate feedback into the system. The disclosed system includes not just the ability to upload entire data sets, but also the ability to upload individual observations over time. Such incremental uploads can then trigger another batch of training, potentially of multiple models with multiple parameter sets resulting in the creation of a new version of the model which could then be selected for use in an application.

The core idea is to describe a system for the on-going management and maintenance of a predictive modeling system that allows training data to be updated through time by the addition (and / or exclusion) of individual observations. Any update of the individual observations that comprise a data set could then trigger (perhaps after being coalesced with other updates to the data set) a complete re-build of the model(s) using configured parameter sets. Such rebuild would be automatic and analogous to the continuous integration systems used...