I am starting my journey in the Machine Learning domain and noticed some tutorials use Excel to explain linear regression. Is it practical to use spreadsheets for building models in a professional setting, or is it strictly for educational purposes before moving onto Python libraries?
3 answers
For anyone in the US working in data science, Excel is mostly considered a "teaching tool" for machine learning. You can certainly use the "Solver" add-in to perform optimization and linear regression, which helps you understand the underlying math. However, it’s not practical for professional production environments because it cannot handle large datasets efficiently and lacks version control. It’s great for a "quick and dirty" proof of concept, but you’ll eventually need to migrate that logic to Python or R.
Have you tried using the Data Analysis Toolpak for your regression models yet, or are you looking to build the formulas manually?
It's strictly educational. No one is deploying an .xlsx file as a predictive API in a real-world tech company in the US.
True, Heather, but it's a great "sanity check" tool. Sometimes I'll run a small sample through Excel just to make sure my Python output isn't hallucinating something wild!
Building them manually using cell formulas is actually a brilliant way to learn how the loss function works. Once you see how the weights update in a spreadsheet, the transition to a library like Scikit-Learn feels much more intuitive. I wouldn't recommend it for a client project, but as a learning milestone, it’s unbeatable. Most senior engineers I know in Silicon Valley actually started exactly this way to grasp the fundamentals.