Mark Daynes - Director of Beyond Systems
As you'd expect from the winners of the Specialized Partner of the Year: Business Analytics at the Oracle UKI Specialized Partner Awards 2014, Beyond work with leading edge BI Applications primarily within the UK Public Sector. We intend to share some of our ideas and discoveries via our blog and hopefully enrich the wider discussion surrounding Oracle Business Intelligence and driving improved insight for customers
Oracle Data Visualizer has been out for a couple of years now and is already on version 4. I'm a big fan and have been digging deep into the latest release which has brought in a substantial amount of changes. They are all available here, but I think that the most exciting inclusions are around the Explain capability and new algorithms that have been included in the product focused on Sentiment Analysis and Machine Learning, as well as the opportunity to load up your own custom scripts.
As an example, let us perform some Sentiment Analysis. I have created some sample data by means of some short reviews of three ficticious restaurants.
Two look pretty good to me and one somewhat less so. Let's push this through the sentiment analyzer and see what results we get. Firstly I navigate to the new super-dynamic Home Page in Data Visualizer v4 and selet the Data tab on the left hand side
As per previous versions, we can upload the data - it can of course be sourced from multiple types of sources, but for this example we're just uploading my small review spreadsheet.
Now we have the data file, we can goto the Data Flows section and create a new data flow. Here we start the flow with the source restaurant review data file.
Note that there are a substantial number of Machine Learning models now available to use in the flow and we will be covering examples of these in further posts.
So, let us add a Sentiment Analysis as the next part of the flow. We will tell Data Visualizer to use the Review column as the source of the analysis and to write out the sentiment to a new column called Emotion.
Let us now add the final storage step to hold the results of the output of the flow. If you look at the table below you can also see that the Sentiment Analysis has done it's job already actually and created what I think look to be pretty accurate results.
We will now save and run the data flow - which will be instant - and then we can look at the results by creaing a simple Project and a visulaisation with a bit of colour.
Personally I think we can now really see the investment in the product coming through and not only is getting so much more powerful, it stilll importantly remains intunitive to use and is a great tool to augment "traditional" BI.Last modified on