def test_drop_duplicates():
cols = ['formula', 'target']
formulae = ['NaCl', 'Al2O3', 'NaCl']
df = pd.DataFrame(columns=cols)
df['formula'] = formulae
df['target'] = range(len(formulae))
output = cmp.generate_features(df, drop_duplicates=False)
out0, out1, out2, out3 = output
assert out0.shape[0] == len(formulae)
output = cmp.generate_features(df, drop_duplicates=True)
out0, out1, out2, out3 = output
assert out0.shape[0] == len(formulae) - 1
def test_sum_feat():
cols = ['formula', 'target']
formulae = ['NaCl', 'Al2O3', 'SiO2']
df = pd.DataFrame(columns=cols)
df['formula'] = formulae
df['target'] = range(len(formulae))
output = cmp.generate_features(df, elem_prop='oliynyk', sum_feat=False)
out0, out1, out2, out3 = output
assert out0.shape[1] == 264
output = cmp.generate_features(df, elem_prop='oliynyk', sum_feat=True)
out0, out1, out2, out3 = output
assert out0.shape[1] == 308
def test_outputs():
cols = ['formula', 'target']
formulae = ['NaCl', 'Al2O3', 'NaCl', 'EsNo', 'BaTiO3', 'GaN', 'Am']
targets = np.random.randn((len(formulae)))
df = pd.DataFrame(columns=cols)
df['formula'] = formulae
df['target'] = targets
output = cmp.generate_features(df)
out0, out1, out2, out3 = output
# check returns are the correct variable type
assert (isinstance(out0, pd.core.frame.DataFrame)
and isinstance(out1, pd.core.series.Series)
and isinstance(out2, pd.core.series.Series)
and isinstance(out3, list))
# check returned targets are equal to originally specified
assert np.allclose(out1, targets, rtol=1e-6, atol=1e-10)
# check returned formulae are equal to originally specified
assert np.all(out2.values == formulae)
# check exotic elements are skipped
assert set(out3) == set(['EsNo', 'Am'])
def test_mini():
cols = ['formula', 'target']
formulae = ['NaCl', 'Al2O3', 'SiO2']
df = pd.DataFrame(columns=cols)
df['formula'] = formulae
df['target'] = range(len(formulae))
output = cmp.generate_features(df, elem_prop='oliynyk', mini=False)
out0, out1, out2, out3 = output
orig_feats = out0.shape[-1]
output = cmp.generate_features(df, elem_prop='oliynyk', mini=True)
out0, out1, out2, out3 = output
new_feats = out0.shape[-1]
assert new_feats < orig_feats
def test_nans():
cols = ['formula', 'target']
formulae = ['NaN']
df = pd.DataFrame(columns=cols)
df['formula'] = formulae
df['target'] = range(len(formulae))
output = cmp.generate_features(df)
out0, out1, out2, out3 = output
assert out0.shape[0] != 0 and out0.shape[1] != 0
def test_extend_features():
cols = ['formula', 'target', 'extra_feature1', 'extra_feature2']
formulae = ['NaCl', 'Al2O3', 'SiO2']
df = pd.DataFrame(columns=cols)
df['formula'] = formulae
df['target'] = range(len(formulae))
df['extra_feature1'] = df['target'] + 0.5
df['extra_feature2'] = df['target'] + 1.5
output = cmp.generate_features(df, extend_features=True)
out0, out1, out2, out3 = output
assert 'extra_feature1' in out0.columns and 'extra_feature2' in out0.columns
def test_all_elem_props():
cols = ['formula', 'target', 'extra_feature1', 'extra_feature2']
formulae = ['C', 'B', 'F', 'V', 'NaCl', 'Al2O3', 'SiO2']
df = pd.DataFrame(columns=cols)
df['formula'] = formulae
df['target'] = range(len(formulae))
df['extra_feature1'] = df['target'] + 0.5
df['extra_feature2'] = df['target'] + 1.5
for elem_prop in ELEM_PROPS:
output = cmp.generate_features(df, elem_prop=elem_prop)
out0, out1, out2, out3 = output
# -*- coding: utf-8 -*-
"""
Created on Thu Apr 9 10:07:00 2020
@author: Steven Kauwe
"""
import pandas as pd
from time import sleep
from cbfv import composition
df = pd.read_csv('test_data_extended_feats.csv')
print('Featurizing DataFrame without extended features')
sleep(1)
output = composition.generate_features(df)
X_cbfv = output[0]
print('Featurizing DataFrame with extended features')
sleep(1)
output = composition.generate_features(df, extend_features=True)
X_extended = output[0]