myyc / kleptes

Interactive data mining goodness

κλέπτης

Some world data mining goodness.

• WHO - pretty good, not super complete, very easy to browse.
• OECD - a complete nightmare. Takes minutes to figure out if a dataset has any data.
• World Bank - lots of stuff, consistent structure, a bit tough to search (will improve on this).
• Eurostat - datasets are complete but the format is horrible. The two functions provided help a lot.
• Yahoo Finance - the usual.

Requirements

• Python 3
• pandas
• Redis (to cache stuff; defaults to localhost, hardcoded)
• requests
• bs4 (used only in one function, don't worry)

It's on PyPI so you can simply pip3 install kleptes.

How do I use this?

It's meant to be used in a Jupyter notebook (or anything interactive, so a Jupyter notebook).

from kleptes import *   # also imports pandas as pd. you'll thank me

who_dims()
who_dims("*indic*")  # supports unix searching because who the fuck knows regexps?

who_dataset("GHO")                # wow results much stuff
who_dataset("GHO", "*suicide*")   # fewer

who_dataset("GHO", "MH_12").head()   # the thing you were looking for


# same-ish stuff for the world bank
wb_inds("*hdi*")        # get indicators like '*hdi*', find one ...
wb_dataset("UNDP.HDI.XD", countries=["mz", "za"])


# and again for eurostat
eus_inds("*empl*")                     # loads of stuff!
eus_inds("*part*time*empl*")           # same
eus_dataset("lfsq_eppga")              # and a dataset


# OECD is a bit more complicated, as the dimension is so complex it deserves
# an object of its own.
oecd_inds("*quarterly*")   # search among indicators (so much shit there)

oecd_dims("QNA")                      # get the dimension of said indicator
oecd_dims("QNA").subject("*gross*")   # figure out possible values of each field
oecd_dataset("qna",
             country=["italy", "fra", "germany", "norw*"],
             subject=["B1_GE"],
             measure="cqr",
             frequency="Q")  # as many kwargs as the dims.


# Yahoo finance is the usual
yf_search("Google")                 # search for a symbol ...
yf_get("GOOGL", days=400)           # get some data

Remember: everything should be searchable.

The exposed functions for now are:

• WHO: who_dataset, who_dim and who_get (very low level).
• OECD: oecd_dataset, oecd_dims, oecd_ind.
• World Bank: wb_inds, wb_dataset and wb_get.
• Eurostat: eus_inds and eus_dataset.
• Yahoo Finance: yh_search and yh_get.

They serve different purposes so despite the name they don't do exactly the same thing (e.g. who_dim and oecd_dims do different stuff).

Everything is cached for a while so you won't be flooding the servers unless you want to.