BrainStation Capstone Project Published!

August 2022

Just published my Capstone project from BrainStation’s Data Science bootcamp, with a wordy title of Picking the Best Predictive Model: A Case Study in Analytics Operationalization. The genesis of the project came from my experiences in industry where there is both a tendency to go for the complex (and expensive) solution first, and a healthy skepticism for Machine Learning from the business folk as well. In my opinion, how you actually choose which models to progress into development and deployment needs to be based on the combination of technical scoring criteria and the relative cost-benefit and projected impact of the model on the business KPIs and associated decision-making processes once deployed.

I found a case study to explore this space in the world of Equipment Reliability. Starting with the simplest solution (do nothing) as a baseline, I built different statistical and machine learning predictive models that can prevent equipment failure while maximizing the equipment’s useful life. I then rationalized each model’s performance in qualitative terms in order to understand if the model would be viable for deployment in the chosen business scenario. Short summary: Both statistical models and ML models are found to be viable for the case study with different levels of impact on the business KPIs.

Coles notes version is here

Full meal deal starts with Part 1 here

Cheers, Craig

Read More