Minitab's most flexible, award-winning and powerful machine learning tool, TreeNet® Gradient Boosting, is capable of consistently generating extremely accurate models.
For those new to TreeNet, it is a powerful implementation of the modern machine learning class of algorithms generally known as Stochastic Gradient Boosting. Developed by Jerome Friedman at Stanford University, the technique is known for its superb predictive accuracy. The secret is in the way a model is built: at each iteration a small tree is added to the current ensemble of trees to correct the combined errors of the ensemble.
Utilizing the variety of the supplied loss functions, the process can be tuned for the specific predictive modeling task, like least squares regression, robust regression, classification, etc. To assist with the model interpretation, TreeNet goes one step further and automatically generates various 2D and 3D plots to explain the nature of dependency of the response variable on the model inputs. The model is flexible enough to automatically discover and incorporate various non-linearities and multi-way interactions. A further set of controls allows the user to fine-tune model interactions to meet specific design objectives.
Our TreeNet modeling engine has a degree of accuracy usually unattainable by a single model or ensembles, like bagging or conventional boosting. Our methodology is not sensitive to data errors and does not require time-consuming data preparation, pre-processing or imputation of missing values. With other methods, data errors can be challenging for conventional data mining and catastrophic for conventional boosting. In contrast, the TreeNet model is immune to such errors as it dynamically rejects data that points too much at variance with the existing model or is contaminated with erroneous target labels.
Avoid conventional trial-and-error or walk-in-the-dark techniques going forward. Our TreeNet modeling engine offers a unique set of insights into your models' inner workings with dependence plots. Our 2D partial dependence plots show the nature of the main effects while our 3D partial dependence plots also include 2-way interactions. Armed with the new insights automatically discovered by TreeNet, you will be able to build highly accurate regression and classification models if needed.
Interaction detection within our TreeNet modeling engine establishes whether interactions of any kind are needed in a predictive model. This system not only helps improve model performance, often dramatically, but also assists in the discovery and use of valuable new insights.