|Anand Jha : Data Science - What vs How||Dec 23, 2017 07:36 am|
I keep seeing this discussion on what skills make a good Data Scientist. I tried to do it through the classic "What vs. How Matrix". Listed down all the key Whats and their relative importance and tried to match it to the Hows - through impact rating of a particular skill on a particular What. What's - Clarity on Business objectives (3), ability to translate a Business objective to Analytics objective (7), translate Analytics objective to choice of models (7), ability to make a judicious choice on models (5), data exploration ability (5), ability to sell analytics results (buy-in) (5), ability to handle big data (3) and productivity (3).
7 - very important, 5 - important, 3 - moderately important
Remember the importance are for a Data Scientist, hence although Business Objective is good to have but not must (some one else can explain you), but Business objective to Analytics Goal is a Data Scientist job (hence 7). Similarly real big data would be required in 10-20% of cases and if you smart enough so called Big Data can be translated to in-database computing like Oracle or parallelization on a single machine. In the same way, think objectively, and see what are your results (simple Sumproduct in excel would do this for you).
My results are (in descending order of importance)
1. knowledge on breadth of Analytics models
2. Depth in understanding of few important models
3. Python and Fundamentals (Statistics/Probability)
4. R and SQL
5. tie on Domain knowledge/MBA education/Java
6. SAS/SPSS/Hive/Spark etc. come as next set of skills
Curious to know what others get.