Data handling is not a linear process, this is not a waterfall. Rather,this is a cycle that feeds on itself. At the top of the cycle is collection. You're acquiring information perhaps from one of these sources that we've described. Then the data moves to pre-processing. So it cornes in as a mass of ones and zéros. We need to begin to normalize it into data models that we can use.
From pre-processing we go to hygiene. Hygiene is a way of extracting some of the noise out of your data, cleaning it up so that you have a more clear signal. Thi is important when you're engaged in data hygiene that you don't throw out your original raw source data because you might be confusing signal and noise. You don't want to throw out the information that you're actually trying to extract out. From hygiene we go to analysis. Now you're doing a first order look at what do I have, what am I working with, can I begin to extract some patterns out of this data? This leads us to visualization. It telling me, what's the information that I can extract from this data.
Visualization then leads us to interprétation, l've decided on what the signal is that's coming out of my data. What does it mean? Applying interprétation is where higher-order data analysis occurs.
Finally, we get to intervention. Once you've acquired your data, pre-processed it, cleaned it, analyzed it, visualized it, developed interprétations of it, you then can begin to say 'OK, what are we going to do with this data'?
- SOURCE : MIT ® Big Data & Social Analysis Certification
