KPMG in conjunction with Data Science Nigeria organized a business analytics hackathon where I was adjudged to have the best submission.
Firstly, my perspective on Data Science is storytelling. Your customers give you some data (Big, Small, Structured and/or Unstructured) with some brief, your job is to make sense and tell some stories out the data. A good storyteller takes you on a journey from one point to another without losing you. That is what I tried to achieve with the KPMG Banking survey data.
The general approach is 1.) Data Setup 2.)Pre-Analytics 3.) Model Development 4.)Evaluation 5.)Post-Analytics
Data Setup entails basic ETL processes. Pre-Analytics involves data exploration where you understand the data and the mindset of the data provider. It helps you to trust the data and trusting the data is fundamental to analytics.
Next is Model Development. This is where Data Science concepts and techniques come into play. From Data Cleansing to Feature Engineering to applying Machine learning Models to the data. The results are in the form of complex data structures like decision trees, vectors, tables and all the statistical jargon.
Evaluation is performed on these results to choose in between alternative outcomes from the applied Machine Learning techniques.
Finally, you conduct Post Analytics which is mainly descriptive analytics but prescriptive in nature. More often than none, CEOs and Business Executives don’t understand DT, ROC, Regression, Gradient Boosting, SVM etc. So it is your job to interpret the outcome in a layman’s language and make recommendations or prescriptions. This is a must.
Normally, I invite cleaners and janitors to my desk and ask them to explain my presentations. If they can understand it, then anybody can. Most CEO don’t have time to interpret your chats so the rule according to Albert Einstein is “Make things as simple as possible, but not simpler”.
Click to see full Presentation.