A fair question many people for organizations creating software is whether they are going to use it themselves. If we are asking someone else to purchase the software, we should “drink our own champagne” and use it for our own business. Indeed, IBM does use IBM SPSS Modeler for a number of use cases – ranging from analyzing customer propensity to buy, customer churn, revenue forecasting, and ad hoc analytics.
I talked recently on a webinar
(replay available through the link) introducing the IBM SPSS Modeler for Data Science Experience. We had a lot of attendees and there was a lot of interest in the new Modeler interface and what we have coming up in both IBM SPSS Modeler for Data Science Experience and our IBM SPSS Modeler desktop offerings (which are also planned to receive the new interface). The webinar provided an opportunity for those listening live to ask questions and we got a lot of great questions and feedback.
I wanted to understand what kinds of topics the participants were most interested in -- and IBM SPSS Modeler is a great way to do so. The way to analyze unstructured data is through text analytics. Text analytics provides a way of transforming the sentences, paragraphs, articles, papers and even books into structured data – that is discrete columns. A common use of text analytics is to add lift to machine learning models – knowing for instance whether a customer is happy with your company can produce better predictions on whether the customer will buy again for you. Many Modeler end users use our text analytics capability found in our SPSS Modeler Premium, SPSS Modeler Gold and SPSS Modeler Subscription, Text Analytics add-on options to gain insight into their data
I first used the text analytics capability to extract concepts. Concepts are basically the important words of grouping of words that are important in a piece of text. IBM SPSS Modeler uses natural language processing to find and extract the important words or phrases that are relevant to the text. For instance, this analysis will discard articles like “a” or “the”. IBM SPSS Modeler comes with pre-built libraries for specific situations. I used the one “Product Satisfaction” for situations where customers are evaluating features of a product. Once I extracted all the concepts, I was then able to create a word cloud using IBM SPSS Modeler’s new visualization capability showing the concepts:
The concepts though were too distinct and not that helpful to my analysis. I wanted to group these together – both based on an automatic approach and my knowledge of the subject. I thus created categories. These are closely related description or opinions. The text analytics feature will built categories based on linguistic analysis. Thus, the software created the category of desktop based on the concepts of “desktop”, “licensed desktop edition” and “modeler desktop”. However I also wanted to create some categories on my own – for instance I wanted to group together mentions of Spark, SparkR and Python as open source.
I then ran the categories through the word cloud graph:
The most discussed questions were on Modeler and DSX – but other topics that came up were around open source, desktop, and integration. I thus was able to use the capabilities of IBM SPSS Modeler to gain insight into this webinar.
Many of you are wondering how to ensure you can get the most value out of IBM SPSS Modeler and other IBM data science software. If so please listen to myself, Julianna Delua (Product Marketer, IBM SPSS Modeler) and Judith Hurwitz (president and CEO of Hurwitz & Associates) discuss “Are you ready for machine learning at scale?”
This gives a lot of great tips and best practices around data science and machine learning. #textanalytics#SPSSModeler