Measuring ML solutions and data readiness
After we define the problem and conclude that ML is a potentially good solution to the problem, we need to set up a way of measuring the problem solution and whether it’s ready for production deployment. For Example 1, we need to have a consensus as to what range is acceptable for the house prediction errors and we can use the ML model in production.
ML model performance measurement
To evaluate the performance of ML solutions, we use ML metrics. For regression models, there are three metrics: mean square error, mean absolute error, and r-square. For classification models, we use the confusion matrix. We will discuss that more in the following chapters.
Is the ML solution ready to be deployed? We need to circle back to the model’s original business goals in the ML problem framing:
- For Zeellow, is predicting a house price with 95% accuracy good enough?
- For Zeellow Mortgage, are we allowed to make a decision...
