Evading fading

I’m going to tell you about the hardest personal trait [at least for us] to be determined. Marketing experts, advertizing agents and even sociologists nearly always operate it. Surely, this trait characterizes an average user well enough. I think if I ask – What way do you identify it? How often do you use it to describe your friends? – you will answer at most – No way and Never! Age. Gender, sexual preferences, political opinions – even a new-comer can mine it processing big data. Just make the corresponding dataset available for him. Probably I shall not be mistaken if I say that all other our algorithms did not require much resources to be developed than we spent attempting to design an algorithm identifying age. We faced this task at the very beginning. Our first client asked to mine women aged 25-30 in his dataset. It cost us so many hours to find a solution! And no effect. Up to now. When negotiating we are always trying to convince that age is not psychological term, – you can easily miss years even looking at the exact person [especially if this person is a woman], – but our advice is in vain. Almost every client has a task involving this parameter. Now during initial negotiations with any potential client we declare that we can determine almost any psychological characteristic and do it more accurately than others. But for age! Here is a bunch of age-related facts we have discovered processing different behavioural datasets: 1. There are two milestones, – graduation from school and retirement, – that relate to real age in many countries. 2. However, behavioural patterns are changed qualitatively upon graduating from school and quantitatively upon retiring. 3. As for post-Soviet region – the age that is hardly identified through data mining is 27. A user who Internet activity corresponds to this age can actually be either 20 or 35. 4. As for English-speaking nations – it is a little bit senior – 32. It well may be that someone succeeded in age identification, we do not know. Although personal orientation, interests, needs are much useful to describe a user and to predict accurately [with some limitation] his personal reaction to a stimulus or his behaviour, SOW will apparently include parameter “age” for a long time. P.S. As far as anyone can die at any age the Grim Reaper seems to misuse Big Data as well )