Packaged as a standalone application, the Inferential Generator transformed regression modeling for Advanced Process Control engineers.
The program integrated rapid database query, data wrangling, visualization, statistical methods for handling outliers, regression generation, and allowed for visualization with automated reporting.
"modeling time was reduced from 1-day/model to 10-minutes/model"
Driven by a python backend, modeling time was reduced from 1-day/model to 10-minutes/model which allowed engineers to explore new model implementation and optimize existing models.
A custom datapull algorithm automated queries, fetching more complete datasets at 10x faster than the traditional method.
A Linear Regression Scikit-Learn class was utilized for model generation, displayed using Matplotlib extensions upon a Tkinter GUI.
Lean Six Sigma outlier handling was implemented while also allowing for the user to view distribution plots and interact with the dataset.
Regression error was displayed to the user with the functionality to drill-down into the potential outliers.
Heat map correlations were displayed for feature set selection and to address collinearity in the multivariable system.
Generalized versions of this functionality is available to the public on my GitHub using data from UCI, but the source code for this program remains confidential.
Documentation is critical!
Previously, the models would live in a black box, with no references to the dataset, modeling process, and system equations.
This program automated report generation to a single PDF file with raw data attached, allowing for further analysis prior to implementation and maintaining a historical record of the work completed.
Nick Belgau 