Automated Data Preparation

This Tech Tip explains how to use the Automated Data Preparation node in IBM SPSS Modeler. In Modeler, users can build machine learning models with node-based, visual programming by selecting nodes from palettes and placing them on the stream canvas to build a stream. The stream represents data flow through operations to a destination, which can be output, a model, or data export. IBM SPSS Modeler offers extensive machine learning and text analysis functionalities, as well as tools for loading, understanding, and transforming data.

The Automated Data Preparation node enhances the efficiency of data preparation for machine learning applications because it identifies and rectifies data issues, eliminates problematic fields, and creates new attributes when necessary. Users have the option to accept automatic changes or customize modifications. This tool significantly improves the speed and reliability of model development and scoring processes.

Users can begin by selecting the Auto Data Prep node from the Field Operations palette and integrating it into the stream. Automated Data Preparation options include balancing speed and accuracy, optimizing for speed, optimizing for accuracy, or creating a custom analysis. Users can manage various aspects of data preparation such as Objectives, Fields, Settings, Analysis, and Annotations. The Settings options has comprehensive tools for field settings, preparing dates and times, excluding low-quality input fields, preparing inputs and targets, constructing and selecting features, and managing field names.

Upon completion of the Automated Data Preparation procedure, users will receive an overview of data transformations, including a field processing summary, recommended predictors, changes to each field, and an action summary. Additionally, users can easily clear any analysis to make further modifications. This technical tip will first review the Automated Data Preparation node settings and then compare the results obtained from automated versus custom settings to demonstrate the node’s functionality.