big data thesis

I guess maybe this should go under /adv/ but I kind of want to talk about how to make an empirical study of big data (is it possible to do something, like an analysis) if I don't really have the knowledge of programming/algorithms. Or where should I start studying these topics....

>>8709720
Take a senior-level stats class at your university. If you fail, you cannot do Big Data.

>>8709720
>>8709761
No fuck you god damn mind readers, I hate that I made a dummy Facebook profile years back with no identifying information and a fake name, and yet I still get emails asking me to add people I talked to in highschool and never even got the email of let alone any contact online.

It's creepy.

>>8709786
*if you don't get an A.

>>8709786
I aced a basic level stat class, but I don't think that really is a benchmark, it is considered to be among the "harder" classes, but once I got a bit into it I saw that it was actually nothing. But it got me interested in this stuff and I was kind of optimistic about my skill.

BUT... I am now that far in my studies that I won't be able to take the kind of class you are talking about, so I will have to find another way...

>>8709787
I know, it gets criticised alot, and ofc it can get creepy, but I really don't want to debate the "ethics" part of it right now, I just want to focus this debate purely on the studies

>>8709720
Fuck off

>>8709816
>>8709720

You are literally wasting your time and money, and will end up dropping out without completion, if you do not have good grasp of the following:
- Bayesian Statistics
- Frequentist Statistics (the type you've probably done a bit of; you want a really good grasp of the intricacies and ins/outs of hypothesis testing)
- Programming- C++, R, Python to begin with, keeping in mind you will be doing the data mining entirely by yourself
- Calculus
- Graph theory if you're looking at social media type stuff

I know /sci/ tend to jump on outsiders and call them retards, but I'm being 100% serious here, you do not have the background. I'm a postgraduate student in this field, part of a larger big data group. I've seen so many stupid kids in your exact position who think they're top shit because they got good marks in a wishy washy degree like commerce and/or marketing (not tryna bag you, but I'm willing to bet that monkeys could pass your degree) and just want to do the get-rich-quick thing and jump into a trendy field. They always, always either get kicked out or fail out almost immediately. People doing this have 4 years of background and rigour over you. If you're really that serious, then do a coursework Masters or another undergraduate degree first. Unless if you're literally Terry Tao (which you aren't), you won't be able to go through all the prerequisite knowledge AND write a thesis in 1-2 years. It's not difficult stuff, just a lot to learn.

>>8710591
You're right that he has way too much to learn to be able to do this, but I disagree on the topics.

There is no need to learn 3 programming languages for one master thesis, strip that down to R cause it's the easiest one. (If you don't care about model evaluation / improving, you could literally get a machine learning algorithm going with 4 lines of code).

No need for bayes stuff for one master thesis as well.

There is not that much need for calculus either in my opinion, if anything linear algebra would probably be better here.

I'd argue that the only things he really needs to make a good master thesis is R data analysis and programming knowledge, one introductory book to machine learning with R, an huge statistics book that goes beyond the basics (at least a little) and yeah maybe a small book in graph theory if he's dealing with social network analysis (but then again he'd need another book on network analysis as well).

If you spend 2 month doing nothing else you can do it, but I'd say it's pretty unrealistic OP. Don't do it.

File: P=NP.png (34KB, 1000x625px) Image search: [Google]

34KB, 1000x625px

ok for BIG DATA here is the key:
Assume independence and identically distributed data and test your hypothesis with a critical value of .25 your shit will get published in Nature.

>>8711150
I mention C++ because many people find their R code is too clunky and takes way too long to run certain things so they write C++ code and wrap R code around that.

>>8710591
>>8711150

Yeah, I was kind of expecting this to be the case. Although I wouldn't mind learning R and some other things, but all in all it sounds like too much work.

But even so, I will still try and do a bit more research, meet with some professors and think things through. Otherwise I will have to find a new topic I guess...