SPSS Modeler

 View Only

Enhancing Predictive Models with Text Analytics

By Jacob Stellon posted Fri May 20, 2022 01:26 PM

The majority of data in a business is in an unstructured text format like surveys, tweets, and press releases. Deriving value from this data can be incredibly time intensive if done manually, and natural language processing can seem to be an arcane discipline of AI. However,  a data scientist or business analyst using SPSS Modeler can extract concepts and sentiment from text data to understand themes or improve predictive models. With it's recent enhancements, the text analytics capabilities of SPSS Modeler allow beginner users to quickly extract value from text while giving expert users the tools to create complex category models in a visual interface. The SPSS Modeler team has overhauled the design and functionality of the Interactive Workbench in the Text Mining node including an optimized user interface and the addition of the Resource Editor for advanced control of linguistic resources. Click here to read more about the Text Analytics capabilities in SPSS Modeler. 

The following demo shows how a SPSS Modeler user can improve a predictive model by incorporating text data and leveraging the Text Analytics features in the product.

In this example a car rental company has built a simple model that predicts a customer's status as active or inactive (a customer is classified as active if they have made a reservation in the last 12 months). The company is currently only using CRM data like gender, relationship status, and age; however, the rental agency also has satisfaction survey data from each customer's last reservation. In the past, the rental agency has read through this data to broadly understand themes, but the volume of survey data is too overwhelming for an employee to do any deeper analysis. Watch the video to see how Text Analytics is used to significantly improve a model built to predict customer churn.