Commit
•
bfbadb9
1
Parent(s):
d90d38c
Upload 10 files
Browse files- ASR_model/RandomForestRegressor.pkl +3 -0
- ASR_model/StandardScaler.pkl +3 -0
- ASR_model/X_train.csv +0 -0
- Barrier_model/RandomForestRegressor.pkl +3 -0
- Barrier_model/StandardScaler.pkl +3 -0
- Barrier_model/X_train.csv +0 -0
- Stability_model/RandomForestRegressor.pkl +3 -0
- Stability_model/StandardScaler.pkl +3 -0
- Stability_model/X_train.csv +0 -0
- model_predict_df.py +150 -0
ASR_model/RandomForestRegressor.pkl
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:af0a9462517c64501ea1ad7370d25aab33e0f53e4e599f57416b99a22f91e22f
|
3 |
+
size 5875427
|
ASR_model/StandardScaler.pkl
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f522e9ec6b3d3b3574cd12986e902c4ab24f3c16594079f892872abf5cc888d9
|
3 |
+
size 4085
|
ASR_model/X_train.csv
ADDED
The diff for this file is too large to render.
See raw diff
|
|
Barrier_model/RandomForestRegressor.pkl
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:595508e8f0f5df559fdee81cf46101a113e39dba95e420d5bd4e7e413f8ec9e1
|
3 |
+
size 5843347
|
Barrier_model/StandardScaler.pkl
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a5f28a8dbe994a0c61007a152b4749d4d33ea310bfc955e945aff5ebe0adeb9d
|
3 |
+
size 4133
|
Barrier_model/X_train.csv
ADDED
The diff for this file is too large to render.
See raw diff
|
|
Stability_model/RandomForestRegressor.pkl
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:369399b4692dce213170d16e1d244b877a5d466ea4677a3eb2d90189e60a076f
|
3 |
+
size 55151315
|
Stability_model/StandardScaler.pkl
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:df4eed3d243c922177f640ca7c8c8150165e311dbd854716997beba8d2b8a801
|
3 |
+
size 2821
|
Stability_model/X_train.csv
ADDED
The diff for this file is too large to render.
See raw diff
|
|
model_predict_df.py
ADDED
@@ -0,0 +1,150 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import pandas as pd
|
3 |
+
import numpy as np
|
4 |
+
import joblib
|
5 |
+
from mastml.feature_generators import ElementalFeatureGenerator, OneHotGroupGenerator
|
6 |
+
from pymatgen.analysis.cost import CostAnalyzer, CostDBElements
|
7 |
+
|
8 |
+
def get_cost(comp_list):
|
9 |
+
ca = CostAnalyzer(costdb=CostDBElements())
|
10 |
+
costs = list()
|
11 |
+
for comp in comp_list:
|
12 |
+
cost = ca.get_cost_per_kg(comp=comp)
|
13 |
+
costs.append(cost)
|
14 |
+
return costs
|
15 |
+
|
16 |
+
def get_stability(df_test):
|
17 |
+
d = 'ASR_model/Stability_model'
|
18 |
+
scaler = joblib.load(os.path.join(d, 'StandardScaler.pkl'))
|
19 |
+
model = joblib.load(os.path.join(d, 'RandomForestRegressor.pkl'))
|
20 |
+
df_features = pd.read_csv(os.path.join(d, 'X_train.csv'))
|
21 |
+
|
22 |
+
features = df_features.columns.tolist()
|
23 |
+
df_test2 = df_test[features]
|
24 |
+
X_stab = scaler.transform(df_test2)
|
25 |
+
stabilities = model.predict(X_stab)
|
26 |
+
|
27 |
+
return stabilities
|
28 |
+
|
29 |
+
def get_barrier(df_test):
|
30 |
+
d = 'ASR_model/Barrier_model'
|
31 |
+
scaler = joblib.load(os.path.join(d, 'StandardScaler.pkl'))
|
32 |
+
model = joblib.load(os.path.join(d, 'RandomForestRegressor.pkl'))
|
33 |
+
df_features = pd.read_csv(os.path.join(d, 'X_train.csv'))
|
34 |
+
|
35 |
+
features = df_features.columns.tolist()
|
36 |
+
X_barrier = df_test[features]
|
37 |
+
X_barrier = scaler.transform(X_barrier)
|
38 |
+
barriers = model.predict(X_barrier)
|
39 |
+
|
40 |
+
return barriers
|
41 |
+
|
42 |
+
def get_asr(df_test):
|
43 |
+
d = 'ASR_model/ASR_model'
|
44 |
+
scaler = joblib.load(os.path.join(d, 'StandardScaler.pkl'))
|
45 |
+
model = joblib.load(os.path.join(d, 'RandomForestRegressor.pkl'))
|
46 |
+
df_features = pd.read_csv(os.path.join(d, 'X_train.csv'))
|
47 |
+
|
48 |
+
features = df_features.columns.tolist()
|
49 |
+
df_test = df_test[features]
|
50 |
+
|
51 |
+
X_ASR = scaler.transform(df_test)
|
52 |
+
|
53 |
+
asrs = model.predict(X_ASR)
|
54 |
+
|
55 |
+
# Get ebars and recalibrate them
|
56 |
+
a = 0.42824232546669644
|
57 |
+
b = 0.36341790743237223
|
58 |
+
errs_list = list()
|
59 |
+
for i, x in X_ASR.iterrows():
|
60 |
+
preds_list = list()
|
61 |
+
for pred in model.model.estimators_:
|
62 |
+
preds_list.append(pred.predict(np.array(x).reshape(1, -1))[0])
|
63 |
+
errs_list.append(np.std(preds_list))
|
64 |
+
asr_ebars = a * np.array(errs_list) + b
|
65 |
+
|
66 |
+
return asrs, asr_ebars
|
67 |
+
|
68 |
+
def process_data(comp_list, elec_list):
|
69 |
+
X = pd.DataFrame(np.empty((len(comp_list),)))
|
70 |
+
y = pd.DataFrame(np.empty((len(comp_list),)))
|
71 |
+
|
72 |
+
df_test = pd.DataFrame({'Material composition': comp_list})
|
73 |
+
|
74 |
+
# Try this both ways depending on mastml version used.
|
75 |
+
try:
|
76 |
+
X, y = ElementalFeatureGenerator(composition_df=df_test['Material composition'],
|
77 |
+
feature_types=['composition_avg', 'arithmetic_avg', 'max', 'min','difference'],
|
78 |
+
remove_constant_columns=False).evaluate(X=X, y=y, savepath=os.getcwd(), make_new_dir=False)
|
79 |
+
except:
|
80 |
+
X, y = ElementalFeatureGenerator(featurize_df=df_test['Material composition'],
|
81 |
+
feature_types=['composition_avg', 'arithmetic_avg', 'max', 'min',
|
82 |
+
'difference'], remove_constant_columns=False).evaluate(X=X, y=y, savepath=os.getcwd(), make_new_dir=False)
|
83 |
+
|
84 |
+
df_test = pd.concat([df_test, X], axis=1)
|
85 |
+
|
86 |
+
elec_cls_0 = list()
|
87 |
+
elec_cls_1 = list()
|
88 |
+
elec_cls_2 = list()
|
89 |
+
elec_cls_3 = list()
|
90 |
+
for elec in elec_list:
|
91 |
+
if elec == 'ceria':
|
92 |
+
elec_cls_0.append(1)
|
93 |
+
elec_cls_1.append(0)
|
94 |
+
elec_cls_2.append(0)
|
95 |
+
elec_cls_3.append(0)
|
96 |
+
elif elec == 'mixed':
|
97 |
+
elec_cls_0.append(0)
|
98 |
+
elec_cls_1.append(1)
|
99 |
+
elec_cls_2.append(0)
|
100 |
+
elec_cls_3.append(0)
|
101 |
+
elif elec == 'perovskite':
|
102 |
+
elec_cls_0.append(0)
|
103 |
+
elec_cls_1.append(0)
|
104 |
+
elec_cls_2.append(1)
|
105 |
+
elec_cls_3.append(0)
|
106 |
+
elif elec == 'zirconia':
|
107 |
+
elec_cls_0.append(0)
|
108 |
+
elec_cls_1.append(0)
|
109 |
+
elec_cls_2.append(0)
|
110 |
+
elec_cls_3.append(1)
|
111 |
+
else:
|
112 |
+
raise ValueError('Invalid electrolyte choice detected. Valid choices are "ceria", "mixed", "perovskite", "zirconia"')
|
113 |
+
|
114 |
+
df_test['Electrolyte class_0'] = elec_cls_0 # ceria
|
115 |
+
df_test['Electrolyte class_1'] = elec_cls_1 # mixed
|
116 |
+
df_test['Electrolyte class_2'] = elec_cls_2 # perovskite
|
117 |
+
df_test['Electrolyte class_3'] = elec_cls_3 # zirconia
|
118 |
+
|
119 |
+
return df_test
|
120 |
+
|
121 |
+
def make_predictions(comp_list, elec_list):
|
122 |
+
|
123 |
+
# Check comp and elec list lengths match
|
124 |
+
assert len(comp_list) == len(elec_list)
|
125 |
+
|
126 |
+
# Process data
|
127 |
+
df_test = process_data(comp_list, elec_list)
|
128 |
+
|
129 |
+
# Calculate the cost of the materials
|
130 |
+
costs = get_cost(comp_list)
|
131 |
+
|
132 |
+
# Get the ML-predicted stability of the materials
|
133 |
+
stabilities = get_stability(df_test)
|
134 |
+
|
135 |
+
# Get the ML-predicted ASR barrier of the materials
|
136 |
+
barriers = get_barrier(df_test)
|
137 |
+
|
138 |
+
df_test['ML pred ASR barrier (eV)'] = barriers
|
139 |
+
asrs, asr_ebars = get_asr(df_test)
|
140 |
+
|
141 |
+
pred_dict = {'Compositions': comp_list,
|
142 |
+
'Electrolytes': elec_list,
|
143 |
+
'Cost ($/kg)': costs,
|
144 |
+
'Stability @ 500C (meV/atom)': stabilities,
|
145 |
+
'ASR barrier (eV)': barriers,
|
146 |
+
'log ASR at 500C (Ohm-cm2)': asrs,
|
147 |
+
'log ASR error (Ohm-cm2)': asr_ebars}
|
148 |
+
|
149 |
+
return pd.DataFrame(pred_dict)
|
150 |
+
|