The main concern of data science, Python has Numpy,scipy,statsmodels,scikit-learn,seaborn,pymc,pandas,keras,lasagne and so on.
What are the features of the popular packages of R languages that Python does not have corresponding packages for?
What other R languages can provide functionality that Python cannot directly implement (user-written algorithms do not count)?
Reply content:
Ggplot2,r inside the most hot bag, python inside only a cottage, the function is not complete. The more popular R-Packet Python basically has a counterpart.
Some outlandish r packages, such as those involved in econometrics/statistics in some niche models, Python does not have a corresponding package, after all, R is designed for academic statisticians and Python belongs to the industry (code farming) language.
These days, statisticians who typically propose a new model or algorithm for a artifice usually attach their own corresponding R package to a published paper and hang it on the Cran. The result is that almost any existing model or algorithm can find the corresponding package on R. (It is another matter if the package is written well)
Econometric economists should be more inclined to encapsulate new models or algorithms into macro-stata or SAS, but the number of economists using R in recent years is also growing.
Not enough Python does not have the corresponding package also does not matter, can call R through the Rpy2 package, just fix the bug will be a little headache. If you are willing to spend time, the same logic, procedures, algorithms, you use any language, r,python,c++ generally can be achieved, is willing to take the time, which language to write up smoothly. For example I have done with Stata rewrite R package glmnet This kind of thing .... Finally it turns out to be completely feasible, but the cost of time is very high ... Those Fortran code ah, Stata that grammar ...
A lot of good R packages or good algorithms generally have Python, allowing users to choose their own familiar language to operate. Of course these are the wheels the party takes time to contribute, to salute them!
But have to say, the black magic aspect R may be stronger, this kind of thing, because if use properly, can multiply efficiency, if use is not good, the code readability is not good.
Like Magrittr.
and PipeR
The pipeline operation brings the operation of similar F # pipes to R, which facilitates many nested function operations. Python has an incomplete implementation, / http Pandas.pydata.org/panda S-docs/stable/whatsnew.html#pipe
, I feel that there is no R inside with smooth.
Similar to Ggplot2.
The syntax, + + + + operation.
And then there are some more interesting things to feel:
Swirl
Use R to learn R.
Htmlwidgets for R
Htmlwidgets:threejs
Individuals prefer Rmarkdown's export of repeatable research documents that feel more convenient than IPython notebook. Caret this machine learns the usual packages, R has, Python if you can have it. A python artifact that allows all r packages to be integrated into the Python framework.
Rpy2
With this, you have R. Magrittr, for R, the revolutionary Pipeline,python does not correspond to the corresponding package is generally seem to have, but it is strange as if the py package is smaller, so the reliability is not very reassuring.
Like what
The shiny in R has a partial problem.
In the field of data science, R Studio is the best IDE, and even Microsoft's VS has to mimic a data sciences model, which shows that the model of the Editor + console + data panel + plot panel is the most suitable for the analysis.
Python is not easy to use Ide,spyder package management and console is not good, Anaconda just put the Spyder into a bag. Rodeo is closest to R Studio, but it doesn't work with Node's front-end cards.
R Good package Everyone said almost, I mention a rvest, with magrittr, functionally barely catch up with the Python crawler, but it is more effort to write, it seems intuitive. Novice may say a few limitations.
Shiny individual sketches may not be lethal. However, the upgraded shiny server is fully capable of dealing with product prototypes for data visualization. Left hand write good R to ETL into the database, right hand can cooperate with Ggplot and polty release. It can be completed and used in half a day.
The point is not which language is better, but you, handy. reshape2!!!!!!!!!!!!!!!!
