Selecting the final machine learning model is not ONLY about selecting the best performing model against the chosen model performance metrics. The ML model also has to meet the business requirements as well. Let me give two examples below.
Looking at Recommender Systems, no customer is going to wait for a recommender system to churn recommendations. Thus it is important that recommender systems can quickly compute and show the results else the business can lose a lot of potential revenue or customer loyalty.
Due to regulations and compliance, machine learning models need to be easily explainable to the regulators otherwise businesses may have to spend a lot of effort and time to explain the "black boxes". Thus this may rule out certain models upfront for instance ensemble or boosting models.
As you can see, data scientist role when it comes to machine learning model training is not just throwing data at different types of models alone. We need to be familiar with the assumptions, pros, and cons of the different models so we can determine the most suitable model for the different use cases. :)
Only with this thought process, can data scientist provide value to their employers. As I always tell my mentees, the stronger that thought process is, the better they are at providing solutions to their employers later on and that will make them valuable and close to indispensable to their employers.
So are you honing your thought process? To read more about that process, I have written another post on it, do check it out here (Building the Thought Process to Solve Business Challenges through Data)
I've just started a contribution page at buymeacoffee.com. If I have been adding value to you, please consider making a donation to let me know I have made an impact. :)