Other menus from Data Analysis describe and are testing : They show how your measurements are distributed, whether two populations differ, and which factors matter.
This Machine Learning menu takes another step forward by building a model, that is, a function that predicts Y based on X, and then uses it to find the best setting.
The menu features four starters that are designed for two very distinct situations, and confusing the two wastes a lot of time:
- You already have the data : a history, a design of experiments you've carried out… and you want to derive a model from it and then optimize it: that's Machine Learning then Optimization in the case of finding a compromise among several Ys;
- You don't have the data yet and each test is expensive: you want the software to tell you which tests to run first to cover the study area, and then to tell you which tests to run next in the region of interest. That is Bayesian Optimization, or Mix Optimization if your factors are proportions of a formulation.
Click on the «Data Analysis» menu, then on «Machine Learning.».
