Empowering machine learning architecture using D3Js.

We all know a saying “A picture is worth a thousand words”. This statement has never been more accurate than in reference to a well thought out data visualization. Using charts or graphs to visualize large amounts of complex data is easier than poring over spreadsheets or reports.First and most important step in machine learning is to understand data. Predictions will not be accurate if they are built without understanding underlying data.  Visualization can help in answering “why model is predicting Class#A and not Class#B”. In this article I will talk about how I have used D3Js along side NN ( see example - here).

Why D3Js?

   There are tools like TableAu, Qlikview that can help in visualizing data. Python and R libraries can be used as well then why D3Js?. Answer is simple – D3Js is the best way to create powerful, interactive visualization at lowest possible cost. 

D3Js is extremely fast, responsive and supports large data sets too for creating dynamic animations in web browsers. Developed visuals can be accessed from mobiles or computer. There are initial efforts required to build foundation. Once that is done then you are set to touch the sky. Thanks to Mr. Mike Bostock for making this powerful library open source.

Example

 Please see sample example here. predictions are made using neural network classification algorithm. D3Js is showing predictions and probability for each class.

D3Js in analytics

 I feel D3Js is a good fit in scoring process and make predictions available to the users.   This can be used in model building phase as well but it might not offer major benefits over native python and R libraries. Real power of this D3Js will be in real time analytics (I will cover this in next posts)