This is a guest post from Bala Deshpande of Simafore. Dr. Deshpande’s has 19 years of experience in using analytical techniques. His first exposure to predictive models and analytics was in the field of biomechanics – in identifying correlations and building multiple regression models to predict muscle forces based on electrical activity in muscles. He began his career as an engineering consultant at EASi Engineering, following which he spent several years analyzing data from automobile crash tests and helping to build safer cars at Ford Motor Company.
You know that predictive analytics has really arrived when there are two articles about it in a single issue of Time magazine (Aug 18, 2014 issue). While it is a rapidly maturing field (according to Gartner’s hype cycle, it is already at its plateau of productivity), there are still some areas where improvements are being worked on.
At the recent RapidMiner World 2014 conference in Boston, the keynote speaker, Dr. Usama Fayyad, showed a couple of illustrations of how even the powerhouse of using data on the internet, a.k.a Google, sometimes misses a trick or two. For example when an article about the riots in Greece includes an ad served up for vacationing in Greek islands! While this is admittedly dated (to 2009), the underlying technology has now evolved further to include sentiment analysis to serve up ads. So when the article includes a lot of text with a negative sentiment such as “riots”, “violence”, “bombing” etc, the algorithm knows that displaying a vacation ad which is (mostly) positive sentiment can be damaging to the brand.
A fascinating fact about search is that most work about analyzing images and videos involves the surrounding text that content providers themselves embed to describe them. Most images contain what is known as the “alt text” tag which is a keyword filled description of the image. So search engines dont really understand the image, the way a human does, but simply use a cluster of keywords or tags to categorize images and videos. However even that is still a powerful way of expressing and manipulating data and the fall out from all of this capture is of course, one facet of big data.
Moving on to more practical applications, we see that a need for utilizing image analysis or the results of image analytics to drive predictive analytics and forecasting. One company develops paper and fabric designs and is actively establishing or tagging their entire inventory of designs so that they can answer several important business questions to accomplish key business objectives. For example, answering questions such as
- what factors on the design influence buyer responses?
- How do color and texture of the design interact with each other to influence the buying process?
- Can we quantify the correlation between geographic location of stores and design attributes (intuitively designers understand there is one)?
- Is there an interplay between position of the designs on the shelves and the design attributes?
All of these questions are very important for optimizing the design and distribution process. Particularly when an industry itself is highly commoditized, predictive analytics offers a way for a business to differentiate itself in the view of their buyers. The buyers here are not the end-users of the products, but the big box retailers who source them from design companies and sell them in-store and online.
Another interesting application of predictive analytics using image analysis is what is called a “visual” market basket analysis. Once we have tagged an image (either manually or using an algorithm), it is easy to develop content based recommenders to provide buyers with recommendations based on their past purchases.
To sum it up, predictive analytics continues to evolve and the ongoing quest is to build computer systems capable of analyzing written language, images and videos based on understanding concepts rather than just keywords.