# -*- coding: utf-8 -*-
# License: BSD-3-Clause
# Author: LKouadio <etanoyau@gmail.com>
"""
Set all dataset.
"""
from warnings import warn
fi=False
_DTAGS=(
"bagoue" ,
"gbalo",
"iris",
"semien",
"tankesse",
"boundiali",
"hlogs",
"nlogs",
"mxs",
"huayuan",
"edis"
)
from .dload import (
load_bagoue ,
load_gbalo,
load_iris,
load_semien,
load_tankesse ,
load_boundiali,
load_hlogs,
load_nlogs,
load_huayuan,
load_edis,
load_mxs,
)
from .gdata import (
make_erp ,
make_ves
)
try :
from ._config import _fetch_data
except ImportError:
warn ("'fetch_data' seems not respond. Use 'load_<area name>'"
" instead.")
else: fi=True
__all__=[
"load_bagoue" ,
"load_gbalo",
"load_iris",
"load_semien",
"load_tankesse",
"load_boundiali",
"load_hlogs",
"load_nlogs",
"load_huayuan",
"fetch_data",
"load_edis",
"load_mxs",
"make_erp" ,
"make_ves",
"DATASET"
]
[docs]
def fetch_data (tag, **kws):
tag = _parse_tags(tag, multi_kind_dataset='bagoue')
func= _fetch_data if fi else None
funcs= (load_bagoue , load_gbalo, load_iris, load_semien, load_tankesse ,
load_boundiali, load_hlogs, load_nlogs, load_huayuan, load_edis,
load_mxs )
funcns = list (map(lambda f: f.__name__.replace('load_', ''), funcs))
if tag in (funcns):
func = funcs[funcns.index (tag)]
return func (tag=tag, data_names=funcns, **kws) if callable (func) else None
fetch_data.__doc__ ="""\
Fetch dataset from `tag`.
A tag corresponds to the name area of data collection or each
level of data processing.
Parameters
------------
tag: str, ['bagoue', 'tankesse', 'semien', 'iris', 'boundiali', 'gbalo']
name of the area of data to fetch. For instance set the tag to ``bagoue``
will load the bagoue datasets. If the `tag` name is following by a suffix,
the later specifies the stage of the data processing. As an example,
`bagoue original` or `bagoue prepared` will retrieve the original data and
the transformed data after applying default transformers respectively.
There are different options to retrieve data such as:
* ['original'] => original or raw data -& returns a dict of details
contex combine with get method to get the dataframe like::
>>> fetch_data ('bagoue original').get ('data=df')
* ['stratified'] => stratification data
* ['mid' |'semi'|'preprocess'|'fit']=> data cleaned with
attributes experience combinaisons.
* ['pipe']=> default pipeline created during the data preparing.
* ['analyses'|'pca'|'reduce dimension']=> data with text attributes
only encoded using the ordinal encoder + attributes combinaisons.
* ['test'] => stratified test set data
Returns
-------
dict, X, y : frame of :class:`~watex.utils.box.Boxspace` object
If tag is following by suffix in the case of 'bagoue' area, it returns:
- `data`: Original data
- `X`, `y` : Stratified train set and training target
- `X0`, `y0`: data cleaned after dropping useless features and combined
numerical attributes combinaisons if ``True``
- `X_prepared`, `y_prepared`: Data prepared after applying all the
transformation via the transformer (pipeline).
- `XT`, `yT` : stratified test set and test label
- `_X`: Stratified training set for data analysis. So None sparse
matrix is contained. The text attributes (categorical) are
converted using Ordianal Encoder.
- `_pipeline`: the default pipeline.
Examples
---------
>>> from watex.datasets import fetch_data
>>> b = fetch_data('bagoue' ) # no suffix returns 'Boxspace' object
>>> b.tnames
... array(['flow'], dtype='<U4')
>>> b.feature_names
... ['num',
'name',
'east',
'north',
'power',
'magnitude',
'shape',
'type',
'sfi',
'ohmS',
'lwi',
'geol']
>>> X, y = fetch_data('bagoue prepared' )
>>> X # is transformed # ready for prediction
>>> X[0]
... <1x18 sparse matrix of type '<class 'numpy.float64'>'
with 8 stored elements in Compressed Sparse Row format>
>>> y
... array([2, 1, 2, 2, 1, 0, ... , 3, 2, 3, 3, 2], dtype=int64)
"""
def _parse_tags (tag, multi_kind_dataset ='bagoue'):
""" Parse and sanitize tag to match the different type of datasets.
In principle, only the 'Bagoue' datasets is allowed to contain a tag
composed of two words i.e. 'Bagoue' + '<kind_of_data>'. For instance
``bagoue pipe`` fetchs only the pipeline used for Bagoue case study
data preprocessing and so on.
However , for other type of dataset, it a second word <kind_of_data> is
passed, it should merely discarded.
"""
tag = str(tag); t = tag.strip().split()
if len(t) ==1 :
if t[0].lower() not in _DTAGS:
tag = multi_kind_dataset +' ' + t[0]
warn(f"Fetching {multi_kind_dataset.title()!r} data without"
" explicitly prefixing the kind of data with the area"
" name will raise an error. In future, the argument"
f" should be '{tag}' instead.", FutureWarning
)
elif len(t) >1 :
# only the multi kind dataset is allowed
# to contain two words for fetching data
if t[0].lower() !=multi_kind_dataset:
tag = t[0].lower() # skip the second word
return tag
from ..utils.funcutils import listing_items_format
_l=[ "{:<7}: {:<7}()".format(s.upper() , 'load_'+s ) for s in _DTAGS ]
_LST = listing_items_format(
_l,
"Fetch data using 'load_<type_of_data|area_name>'like",
" or using ufunc 'fetch_data (<type_of_data|area_name>)'.",
inline=True , verbose= False,
)
_DDOC="""\
WATex dataset is composed of different kind of data for software implementation.
- ERP data found in 'gbalo', 'boundiali' localities in northern part of
Cote d'Ivoire <'https://en.wikipedia.org/wiki/Ivory_Coast'>'
- VES data collected in 'gbalo', 'semien', 'tankesse' in center and
eastearn part of Cote d'Ivoire'.
- FLOW RATE FEATURES data computed from Bagoue ERP and VES data.
Refer to paper :doi:`https://doi.org/10.1029/2021wr031623`.
- COMMON MACHINE LEARNING popular data sets such IRIS.
- EDI datasets from huayaun county in China, Hunan Province.
- HLOGS and MXS data for Hongliu coal mine raw and preprocessed borehole data.
"""
DATASET= type ("DATASET", (), {"KIND": _DTAGS,
"HOW":_LST,
"DOC":_DDOC,
}
)
