The party has started in the analytics jungle and I am a little late to join. But here I come.
History of IBM Watson
IBM Watson has been quite a discussed topic in the history of American TV and created a lot of buzz in the computer science world when it famously won the live TV show Jeopardy! in 2011. Watson is a supercomputer specialized in natural language process which on the show, was asked questions like this:
This 1959 Daniel Keyes novella about Charlie Gordon & a smarter-than-average lab mouse won a Hugo Award :
To which Watson correctly answered : Flowers for Algernon
Impressive; isn't it! Well, winning $1M on a show ($500K of which was donated by IBM) and being famous worldwide was one thing, soon people began to talk about how such a powerful system could be put in the application to solve real world problems. Let's fast forward to 5 years into the future: 2016!
Here Comes the IBM Watson Analytics
Me: What are drivers of call volumes at my call center?
Watson Analytics suggests that Time of Day and Weekday has 66% predicting power to explain call volumes. And that's actually true. In my dataset, call volumes heavily depend upon
- Time of Day - lots of call in morning and evening office hours, less calls in sleeping hours
- Weekday - There are lots of call on weekdays and very less calls on weekends
So far so good. I did not need to write a single line of code, did not require to run any regression model, did not even created any chart and I got these amazing insights. Just by typing a question in IBM WA. Yeah! THAT is IBM Watson Analytics.
IBM Watson Analytics is cloud based 'powerful' data analytics platform which connects the power of predictive analytics with natural language processing and lets you ask business questions and answers those questions just on a single click. You can directly import data from Twitter in the real-time and analyse sentiments of public as the event unfolds - all without needing to write a single line of code.
How Much Does WA Cost
Free for a month in the trial period, $30 per user per month afterwards for a plus version, $80 for the professional version. More here on WA website.
How Watson Analytics Works
You can upload Excel or CSV to Watson Analytics or directly connect with your databases (WA currently supports many popular databases, more on this later). Also, you can directly import data from Twitter and upload it on WA cloud storage before you begin your analysis. Once your data has been uploaded in the cloud, you can start asking direct questions to WA and in response (if your question is valid and relevant to WA) WA provides you with data visualizations that best answers your questions. You can further deep dive into those charts and can further explore data either by asking more questions or can create your own analysis. These analysis are called 'discoveries' in Watson Analytics.
Communicating results is the important aspect of knowledge discovery and WA provides you with a couple of options like sending email, tweeting the results, downloading an image/pdf/powerpoint or sending a link to your colleague.
Watson Analytics if powered by several existing key concepts like
- Natural Language Processing
- Cloud Computing
- Predictive Analytics
- Data Visualization
This is merely the surface that I have scratched on Watson Analytics. At this stage, I am kind of fascinated and fancied and I am yet to experience how it can be accepted, adopted and add value in a real production environments by real business users. I will share more in the future posts as I explore topic further in depth.
I would love to know your experience with WA so far in comments below.