•1 min read•from Data Science
Why is it that stakeholders expect ML models to have 0% error rate?
Definitely the most frustrating thing as working as a Data Scientist. You run an experiment, find that building a model greatly increase metric X at almost no cost, has safe model metrics, present it to stakeholders, everybody agrees with proceeding to deploying and utilizing the model in production, and yet every time the model takes a wrong decision, we get questioned about it. Why did the model say this?
Man when did I ever say the model obtained a 100% accuracy in the validation phase? Why is it so hard for stakeholders to understand that the best models humankind ever created are expected to make wrong calls once in a while?
[link] [comments]
Want to read more?
Check out the full article on the original site
Tagged with
#ML models
#Stakeholders
#Data Scientist
#Model accuracy
#Validation phase
#Error rate
#Production deployment
#Model metrics
#Experiment
#Model performance
#Machine Learning
#Metric X
#Wrong decisions
#Model utilization
#Model development
#Data Science
#Model evaluation
#100% accuracy
#Model risk
#Model calibration