Analytics
News, events, thought leadership and more.
At Alteryx we strive to achieve a balance of analytic sophistication and ease of use with all of our tools, including our predictive functions. The first predictive tools in Alteryx Analytic arrived way back in version 7.0 circa June 2012 and have served Alteryx users well in that time, earning a “Visionary” rating in Gartner’s 2016 Magic Quadrant for Advanced Analytics Platforms, and is now a “Challenger” in Gartner’s 2017 Magic Quadrant for Data Science Platforms. But things move quickly in the data science community, machine learning algorithms evolve to become more and more robust, and analysts demand more sophisticated measures to understand the performance of models.
Alteryx Analytics11.0 brings key enhancements to three of the most commonly used predictive models: Linear Regression, Logistic Regression, and Decision Tree. Linear Regression and Logistic Regression now support elastic net regularization, a technique to prevent overfitting – or creating too complex of a model that doesn’t generalize well to data outside of your training set. The Decision Tree tool now supports C5.0 – a modern decision tree algorithm that can create more accurate models for classification problems. All three tools now support cross-validation so that you can get unbiased performance metrics from the reports without having to run the model across a separate validation dataset.
With Alteryx Analytics 11.0, all three of these tools display a unified configuration interface and brand new interactive model reports. The new configuration highlights only the required parameters (model name, target variable, and predictor variables), and pushes all other configuration to a customization area, keeping the output clean and focused on what’s most critical. The reports contain all the information a typical data scientist would look for (R Squared, MAPE, etc) but also contain measures and plots that anyone can understand like variable importance (impact) and the Predicted vs Actual chart (see below). These enhancements are meant to speed the model development lifecycle for both experienced data scientists as well as those trying out the predictive tools for the first time.
The predictive enhancements we’ve made are just one of the many new features we’ve released in 11.0. Make sure to attend the live webinar, “Bridging the Gap: Self-Service Data Analytics meets Data Governance” on March 8th to see this and the other new features in 11.0 in action. You can also, learn more about the new features in Alteryx Analytics 11.0 by following our release blog series.
