Machine Learning Overview (page 3 of 13)

All trained machine learning models have error. In our machine learning processes, we have two broad categories of errors: 1) reducible (minimizable) and 2) irreducible (cannot be reduced or minimized due to the stochastic nature of any business problem involving human behavior). Business analysts attempt to reduce the reducible error component by (among others) adding new data (more rows), adding new features (more columns), managing outliers, modifying hyper-parameters or using different algorithms. Irreducible error comes from the fact that all phenomenon have a random component that is not predictable using any algorithm.