This update includes additional visualizations, integrations and general improvements.
Updated Visualization
Variables for Binned Scatterplot, Parallel Coordinates Plot, and Heat Map
User Benefit
Graphs can be split by variables faster, removing the need to first subset the data before creating these graphs.
Summary
In each of these dialogs, grouping variables can be designated using the By Variable section.
Improvement: Lag
User Benefit
Lagged columns are commonly used in time series modeling and supervised machine learning, such as CART, TreeNet, and Random Forests. This improvement allows for faster data preparation.
Summary
Numerous lag columns can be easily generated for one or multiple time series columns. This command is located in Stat-Time Series-Lag.
Addition
R Integration
Call R scripts from Minitab Statistical Software. R is a language and environment for statistical computing and graphics.
R scripts can run in 3 ways:
- Run the RSCR in the command line pane
- Run a Minitab exec that includes the RSCR command
- Customize the interface to run a Minitab exec that includes the RSCR command
Summary
Execute external R scripts that use Minitab Statistical Software variables (columns, constants, matrices) as inputs. Results are returned to Minitab and displayed in the output navigator and output pane.
Addition
Minitab’s Predictive Analytics Module
User Benefit
Tree-based methods empower predictive analytics with not only speed to answer, but also remarkable accuracy and ease of interpretation. Users can quickly understand the key drivers of a process.
Summary
Our proprietary, best-in-class, tree-based machine learning algorithms not only have the power to provide deeper insights and visualize multiple complex interactions with decision trees but are equipped to handle larger data sets with more variables, messy data, missing values, random outliers, and non-linear relationships. These methods are now available in a module that you easily add to Minitab Statistical Software.
Addition
Random Forests®
New Feature: Random Forests consists of many individual decision trees that operate as an ensemble.
User Benefit
Random Forests generally provides better predictive power than a single decision tree.
Summary
Based on a collection of CART Trees, Random Forests leverages repetition, randomization, sampling, and ensemble learning in one convenient place that brings together independent trees and determines the overall prediction of the forest.
Addition
TreeNet® (Gradient Boosting)
New Feature: TreeNet Classification and TreeNet Regression. Includes Fit Model and Discover Key Predictor submenus.
User Benefit
Gradient boosting can deliver optimal prediction accuracy and unique insights.
Summary
Our most flexible, award-winning and powerful machine learning tool, TreeNet Gradient Boosting, is known for its superb and consistent predictive accuracy due to its iterative structure that corrects combined errors of the ensemble as it builds.
New Visualizations
Correlogram
User Benefit
Correlograms are useful for finding important correlations when faced with many variables. Viewing correlations as a color gradient is an alternative approach to displaying a matrix plot or a table of correlation statistics.
Summary
The correlogram makes it easy to visualize a matrix or correlations, particularly when the number of variables is large.
