I work in a financial firm in the IT department and have noticed we have a bunch of quant analysts, statisticians, etc. They all make their own custom programs and models in Python. I asked them why and they said it's what everybody uses for big data analytic. Any specific reason why Python seems to be superior language in this, or is it just a meme? Should I learn Python?
It's just easy to write, easy to read, popular, and has lots of packages.
If you want to do the kind of analysis that they do, then learning the language they use seems like something you shouldn't need to ask us for advice about. If you have another purpose in mind, then the anecdote you gave seems irrelevant.
>>51700627
I know very little about it, but it seems to be the language with all of the relevant libraries and tutorials for data science/ML.
Why all the libraries and tutorials are in python? Not sure...
AFAIK it's because of libraries (SciPy, pandas), but i don't know why they prefer python over R
>>51700659
between the two, Python's enormously more readable and intuitive.
R is also surprisingly inefficient with larger datasets, which doesn't help R's case, but Python (while not being super efficient itself) at least lends itself really well to interoperability with other languages (like hooking into C) for the stuff you need to run more efficiently.
>>51700649
I was thinking of creating my own person budget program with it. Create categories, have a deposit and withdraw option, and it automatically keeps track of my NET balance, pie chart of my spending, etc.
>>51700627
Because it's a rather powerful language and it's easily accessible to people with little or no prior exposure to programming. The easy syntax lets them express complicated calculations easier than say in C/C++ which arguably would be better by performance, but you know... it looks like fucking sorcery to normal human beings.
PS. Afaik, statistics is done mostly in R. At least where I'm at the university courses are based on that.
>>51700682
So your own shitty version of the account management that your bank offers? Or something to aggregate myriad accounts, like a shitty version of Mint?
I'm not saying that this is a complete waste of time, but this would be an exercise, not a real task. I can guarantee that a novice programmer isn't going to build anything more comprehensive or useful than the stuff that's already out there.
Use Mint or whatever, take notes about whatever features you wish were available but can't find, then start thinking about how you could code those features up (especially by connecting with Mint or whatever).
If you're just looking for exercises to get your head into the mindset of programmatically solving problems, try Project Euler. It's mathy, but that's what programming languages are good at.
>>51700771
>but this would be an exercise, not a real task. I can guarantee that a novice programmer isn't going to build anything more comprehensive or useful than the stuff that's already out there.
How the fuck else do you learn?
>>51700659
R is great at data analysis when performance is not to big of a concern. R are does some stupid shit like not being able to append vectors, and it has to re-initialize objects in a for loop. Little shit like that makes R not as good as Python for production analytics.
You really should learn both Python and R. R is nice because it abstracts away almost all the complexity of predictive modelling and in my opinion is easy to pick up. Python is better though if you need your stuff to run faster.
>>51700799
>How do you learn?
By starting with something less ambitious than a banking system. Like a problem on Project Euler, which I specifically suggested for someone looking for exercises for learning.
And by being moderately respectful to people who offer you advice.
>>51700799
>How the fuck else do you learn?
from his post...
>If you're just looking for exercises to get your head into the mindset of programmatically solving problems, try Project Euler. It's mathy, but that's what programming languages are good at.
if you've never programmed before, something to track all of your financial accounts would be insane.
>>51700983
>not being able to append vectors
What
>vec = c(vec, x)
>>51701576
You are creating a new vector and the old one still stays bloating your memory.
>>51701692
this. it makes sense on some level because somewhere someone decided you don't want to lose the old data frame, but that's not the paradigm anyone needs anymore.
>>51700649
This. Theres a plethora of tools and modules out there for analyzing huge data sets.
>>51700682
Python can do that easy peasy. Use pyodbc to connect to a sql backend and your set. Or store data to a file i guess. Lots of modules already out there for making simple charts /graphs
do any of you work as a data analyst? I took a data mining course this semester and intend to pursue a master's in data analytics here. I've gotten well versed in R and intend to work some more with Python and SQL over winter break. What else is common to use besides shit like SAS
>>51703089
or just use SQLite or BerkeleyDB instead of just dumb ass flat files or a need for a db daemon.
Python is easy to read and write, so non-coders love it.
>>51703236
easy if you use other people's code I guess.
feels incredibly C like in many ways.
>>51700627
It's either R or Python, anon. I prefer R, the syntax is a little more intuitive for me, since I don't come from a comp sci/programming background.
Easy to learn, easy to understand.
You can do things in Python in a few lines that would require writing a bunch of new functions and typedefs and debugging for hours in c*.
