Keywords: Tableau, Alteryx, R, Utilities, Natural Gas, Energy, Forecast, What-If Analysis, XGBoost, Multivariate Statistics
Utility data has certain qualities that are conducive to strong results from machine learning. In this session we talk through a specific use case from proof-of-concept to results and model deployment while discussing how we trained employees to use the new software, worked with consultants to fill in skill gaps, and convinced business to use the tool we created.
How much revenue is lost each year to advances in appliance efficiency? What is our financial exposure to variances in the weather? How do we enact a self-service BI initiative at a natural gas utility? How do we get business to actually use the product of our research?
Powered by Alteryx and R, we show an interactive table in Tableau complete with a Multivariate F-statistic quantifying relative rarity of each month's consumption behaviors. In one Tableau workbook we visualized the machine learning retrain results and a 12 month what-if analysis/forecast of usage by rate class.
Session attendees will learn to use a simple for-loop in R to generate a multivariate statistic to enhance any data table. We will show how Alteryx, R, and Tableau can be used in tandem to query data, automate machine learning retraining, and visualize results for end users.