Business Problem or Pain Point
The City of Tallahassee, Florida wanted to know if any of their 400+ commercial water meters in the city were about to fail. This ideally should be known in advance, before sending a technician to inspect it. This would help reduce maintenance costs and save labor costs on the technicians.
Using the water meter sensor data, we built a highly accurate AI model which predicted if a water meter will fail in advance. Each water meter got assigned a risk-of-failure score, and the scores were displayed on a dashboard. From this dashboard, the city could export a list of meters which had the highest risk-of-failure that required preventative maintenance. The lists were provided to technicians to do an inspection of the meters. Our AI model achieved a 233% higher detection rate versus the baseline detection rate from their random inspections.
Technologies and Algorithms Used
R, Alteryx, Shiny, AutoML, Gradient Boosting
Remix Institute is a technology firm that focuses on data science, machine learning, and AI for business via services, training, products, and software.