-
Notifications
You must be signed in to change notification settings - Fork 117
/
sklearn2pmml.py
399 lines (313 loc) · 14.8 KB
/
sklearn2pmml.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
import sys
from pandas import DataFrame, Series
from sklearn_pandas import DataFrameMapper
from sklearn.ensemble import GradientBoostingClassifier, GradientBoostingRegressor, RandomForestClassifier, RandomForestRegressor
from sklearn.linear_model import LinearRegression, LogisticRegression
from sklearn.neural_network import MLPRegressor
from sklearn.pipeline import make_pipeline
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import KBinsDiscretizer, StandardScaler
from sklearn.tree import DecisionTreeClassifier
from sklearn2pmml.cross_reference import make_memorizer_union, make_recaller_union
from sklearn2pmml.cross_reference import Memory, Memorizer, Recaller
from sklearn2pmml.decoration import Alias, CategoricalDomain, ContinuousDomain, MultiAlias
from sklearn2pmml.ensemble import EstimatorChain, GBDTLMRegressor, GBDTLRClassifier, Link, SelectFirstClassifier, SelectFirstRegressor
from sklearn2pmml.expression import ExpressionClassifier, ExpressionRegressor
from sklearn2pmml.neural_network import MLPTransformer
from sklearn2pmml.pipeline import PMMLPipeline
from sklearn2pmml.postprocessing import BusinessDecisionTransformer
from sklearn2pmml.preprocessing import CutTransformer, SelectFirstTransformer
from sklearn2pmml.ruleset import RuleSetClassifier
from sklearn2pmml.tree.chaid import CHAIDClassifier, CHAIDRegressor
from sklearn2pmml.util import Expression, Predicate
import numpy
from common import *
sys.path.append("../")
from main import *
datasets = []
if __name__ == "__main__":
if len(sys.argv) > 1:
datasets = (sys.argv[1]).split(",")
else:
datasets = ["Audit", "Auto", "Housing", "Iris", "Versicolor", "Wine"]
def make_bins(X, cols):
result = dict()
for col in cols:
bins = numpy.nanquantile(X[col], q = [0.0, 0.25, 0.50, 0.75, 1.0], interpolation = "nearest")
# Deduplicate
bins = list(dict.fromkeys(bins))
# Unbox Numpy float scalars to Python floats
bins = [float(bin) for bin in bins]
result[col] = bins
return result
def make_bin_labels(bins):
result = dict()
for key, value in bins.items():
result[key] = list(range(0, len(value) - 1))
return result
def make_chaid_dataframe_mapper(cont_cols, cat_cols, bins = {}, labels = {}):
return DataFrameMapper(
[([cont_col], [ContinuousDomain(), CutTransformer(bins = bins[cont_col], labels = labels[cont_col])]) for cont_col in cont_cols] +
[([cat_col], [CategoricalDomain()]) for cat_col in cat_cols]
, df_out = True)
def build_chaid_audit(audit_df, name):
audit_X, audit_y = split_csv(audit_df)
cont_cols = ["Age", "Hours", "Income"]
cat_cols = ["Education", "Employment", "Marital", "Occupation", "Gender"]
bins = make_bins(audit_X, cont_cols)
labels = make_bin_labels(bins)
pipeline = PMMLPipeline([
("mapper", make_chaid_dataframe_mapper(cont_cols, cat_cols, bins, labels)),
("classifier", CHAIDClassifier(config = {"max_depth" : 9, "min_child_node_size" : 10}))
])
pipeline.fit(audit_X, audit_y)
store_pkl(pipeline, name)
node_id = Series(pipeline._final_estimator.apply(audit_X), dtype = int, name = "nodeId")
store_csv(node_id, name)
if "Audit" in datasets:
audit_df = load_audit("Audit", stringify = False)
build_audit(audit_df, GBDTLRClassifier(RandomForestClassifier(n_estimators = 17, random_state = 13), LogisticRegression()), "GBDTLRAudit")
audit_df = load_audit("Audit")
build_multi_audit(audit_df, EstimatorChain([("gender", Link(DecisionTreeClassifier(max_depth = 5, random_state = 13), augment_funcs = ["predict_proba", "apply"]), str(True)), ("adjusted", LogisticRegression(), str(True))]), "MultiEstimatorChainAudit")
build_chaid_audit(audit_df, "CHAIDAudit")
audit_df = load_audit("AuditNA")
build_chaid_audit(audit_df, "CHAIDAuditNA")
def _versicolor_func(petal_length, petal_width):
if petal_length > 2.45:
if petal_width <= 1.75:
return 2.2824 # logit(49. / 54.)
else:
return -3.8067 # logit(1. / 46.)
else:
return -9.2102 # logit(0.0001)
def build_expr_versicolor(versicolor_df, name):
versicolor_X, versicolor_y = split_csv(versicolor_df)
class_exprs = {
1: Expression("_versicolor_func(X['Petal.Length'], X['Petal.Width'])", function_defs = [_versicolor_func])
}
pipeline = PMMLPipeline([
("classifier", ExpressionClassifier(class_exprs, normalization_method = "logit"))
])
pipeline.fit(versicolor_X, versicolor_y)
store_pkl(pipeline, name)
species = DataFrame(pipeline.predict(versicolor_X), columns = ["Species"])
species_proba = DataFrame(pipeline.predict_proba(versicolor_X), columns = ["probability(0)", "probability(1)"])
species = pandas.concat((species, species_proba), axis = 1)
store_csv(species, name)
if "Versicolor" in datasets:
versicolor_df = load_versicolor("Versicolor")
build_versicolor(versicolor_df, GBDTLRClassifier(GradientBoostingClassifier(n_estimators = 11, random_state = 13), LogisticRegression(solver = "liblinear")), "GBDTLRVersicolor")
build_expr_versicolor(versicolor_df, "ExpressionVersicolor")
def build_chaid_iris(iris_df, name):
iris_X, iris_y = split_csv(iris_df)
mapper = DataFrameMapper([
(iris_X.columns.values, KBinsDiscretizer(n_bins = 5, encode = "ordinal", strategy = "kmeans"))
])
pipeline = PMMLPipeline([
("mapper", mapper),
("classifier", CHAIDClassifier(config = {"max_depth" : 3, "min_child_node_size" : 10}))
])
pipeline.fit(iris_X, iris_y)
store_pkl(pipeline, name)
node_id = Series(pipeline._final_estimator.apply(iris_X), dtype = int, name = "nodeId")
store_csv(node_id, name)
def _iris_setosa_func(petal_length):
if petal_length <= 2.45:
return 1
else:
return 0
def _iris_versicolor_func(petal_length, petal_width):
if petal_length > 2.45:
if petal_width <= 1.75:
return 0.90741
else:
return 0.02174
else:
return 0
def _iris_virginica_func(petal_length, petal_width):
if petal_length > 2.45:
if petal_width <= 1.75:
return 0.09259
else:
return 0.97826
else:
return 0
def build_expr_iris(iris_df, name):
iris_X, iris_y = split_csv(iris_df)
class_exprs = {
"setosa" : Expression("_iris_setosa_func(X['Petal.Length'])", function_defs = [_iris_setosa_func]),
"versicolor" : Expression("_iris_versicolor_func(X['Petal.Length'], X['Petal.Width'])", function_defs = [_iris_versicolor_func]),
"virginica" : Expression("_iris_virginica_func(X['Petal.Length'], X['Petal.Width'])", function_defs = [_iris_virginica_func])
}
pipeline = PMMLPipeline([
("classifier", ExpressionClassifier(class_exprs, normalization_method = "simplemax"))
])
pipeline.fit(iris_X, iris_y)
store_pkl(pipeline, name)
species = DataFrame(pipeline.predict(iris_X), columns = ["Species"])
species_proba = DataFrame(pipeline.predict_proba(iris_X), columns = ["probability(setosa)", "probability(versicolor)", "probability(virginica)"])
species = pandas.concat((species, species_proba), axis = 1)
store_csv(species, name)
def build_mlp_iris(iris_df, name, transformer, predict_proba_transformer = None):
iris_X, iris_y = split_csv(iris_df)
pipeline = PMMLPipeline([
("decorator", ContinuousDomain()),
("transformer", transformer),
("classifier", LogisticRegression(random_state = 13))
], predict_proba_transformer = predict_proba_transformer)
pipeline.fit(iris_X, iris_y)
pipeline.verify(iris_X.sample(frac = 0.05, random_state = 13))
store_pkl(pipeline, name)
species = DataFrame(pipeline.predict(iris_X), columns = ["Species"])
if predict_proba_transformer is not None:
species_proba = DataFrame(pipeline.predict_proba_transform(iris_X), columns = ["probability(setosa)", "probability(versicolor)", "probability(virginica)", "decision"])
else:
species_proba = DataFrame(pipeline.predict_proba(iris_X), columns = ["probability(setosa)", "probability(versicolor)", "probability(virginica)"])
species = pandas.concat((species, species_proba), axis = 1)
store_csv(species, name)
def build_pca_iris(iris_df, name):
iris_X, iris_y = split_csv(iris_df)
pca_names = ["pca_1", "pca_2"]
memory = Memory()
pca_union = FeatureUnion([
("first", ExpressionTransformer("X[0]")),
("second", ExpressionTransformer("X[1]")),
("ratio", Alias(ExpressionTransformer("X[0] / X[1]"), name = "pca_ratio", prefit = True))
])
pipeline = PMMLPipeline([
("pca", make_pipeline(MultiAlias(PCA(n_components = len(pca_names)), names = pca_names), make_memorizer_union(memory = memory, names = pca_names))),
("classifier", LogisticRegression())
], predict_transformer = make_pipeline(make_recaller_union(memory = memory, names = pca_names), pca_union))
pipeline.fit(iris_X, iris_y)
store_pkl(pipeline, name)
species = DataFrame(pipeline.predict_transform(iris_X), columns = ["Species", "xref({})".format(pca_names[0]), "xref({})".format(pca_names[1]), "pca_ratio"])
store_csv(species, name)
def build_ruleset_iris(iris_df, name):
iris_X, iris_y = split_csv(iris_df)
classifier = RuleSetClassifier([
(Predicate("X['Petal.Length'] >= 2.45 and X['Petal.Width'] < 1.75"), "versicolor"),
("X['Petal.Length'] >= 2.45", "virginica")
], default_score = "setosa")
pipeline = PMMLPipeline([
("domain", ContinuousDomain(display_name = ["Sepal length (cm)", "Sepal width (cm)", "Petal length (cm)", "Petal width (cm)"])),
("classifier", classifier)
])
pipeline.fit(iris_X, iris_y)
pipeline.verify(iris_X.sample(frac = 0.10, random_state = 13))
store_pkl(pipeline, name)
species = DataFrame(pipeline.predict(iris_X), columns = ["Species"])
store_csv(species, name)
def build_selectfirst_iris(iris_df, name):
iris_X, iris_y = split_csv(iris_df)
pipeline = PMMLPipeline([
("mapper", DataFrameMapper([
(iris_X.columns.values, ContinuousDomain())
])),
("classifier", SelectFirstClassifier([
("select", Pipeline([
("classifier", DecisionTreeClassifier(random_state = 13))
]), "X[1] <= 3"),
("default", Pipeline([
("scaler", StandardScaler()),
("classifier", LogisticRegression(multi_class = "ovr", solver = "liblinear"))
]), str(True))
]))
])
pipeline.fit(iris_X, iris_y)
store_pkl(pipeline, name)
species = DataFrame(pipeline.predict(iris_X), columns = ["Species"])
species_proba = DataFrame(pipeline.predict_proba(iris_X), columns = ["probability(setosa)", "probability(versicolor)", "probability(virginica)"])
species = pandas.concat((species, species_proba), axis = 1)
store_csv(species, name)
if "Iris" in datasets:
iris_df = load_iris("Iris")
build_chaid_iris(iris_df, "CHAIDIris")
build_expr_iris(iris_df, "ExpressionIris")
mlp = MLPRegressor(hidden_layer_sizes = (11, ), solver = "lbfgs", random_state = 13)
build_mlp_iris(iris_df, "MLPAutoencoderIris", MLPTransformer(mlp), predict_proba_transformer = Alias(BusinessDecisionTransformer("'yes' if X[1] >= 0.95 else 'no'", "Is the predicted species definitely versicolor?", [("yes", "Is versicolor"), ("no", "Is not versicolor")], prefit = True), "decision", prefit = True))
build_mlp_iris(iris_df, "MLPTransformerIris", MLPTransformer(mlp, transformer_output_layer = 1))
build_pca_iris(iris_df, "PCAIris")
build_ruleset_iris(iris_df, "RuleSetIris")
build_selectfirst_iris(iris_df, "SelectFirstIris")
def build_chaid_auto(auto_df, name):
auto_X, auto_y = split_csv(auto_df)
cont_cols = ["acceleration", "displacement", "horsepower", "weight"]
cat_cols = ["cylinders", "model_year", "origin"]
auto_X[cat_cols] = auto_X[cat_cols].astype("Int64")
bins = make_bins(auto_X, cont_cols)
labels = make_bin_labels(bins)
pipeline = PMMLPipeline([
("mapper", make_chaid_dataframe_mapper(cont_cols, cat_cols, bins, labels)),
("regressor", CHAIDRegressor(config = {"max_depth" : 7, "min_child_node_size" : 10}))
])
pipeline.fit(auto_X, auto_y)
store_pkl(pipeline, name)
node_id = Series(pipeline._final_estimator.apply(auto_X), dtype = int, name = "nodeId")
store_csv(node_id, name)
def _scale_displacement(displacement):
return (displacement - 194.41) / 104.51
def _scale_weight(weight):
return (weight - 2977.58) / 848.32
def build_expr_auto(auto_df, name):
auto_X, auto_y = split_csv(auto_df)
expr = Expression("-1.724 * _scale_displacement(X['displacement']) + 4.879 * _scale_weight(X['weight']) + 23.45", function_defs = [_scale_displacement, _scale_weight])
pipeline = PMMLPipeline([
("regressor", ExpressionRegressor(expr, normalization_method = "none"))
])
pipeline.fit(auto_X, auto_y)
store_pkl(pipeline, name)
mpg = Series(pipeline.predict(auto_X), name = "mpg")
store_csv(mpg, name)
if "Auto" in datasets:
auto_df = load_auto("Auto")
# XXX
auto_df["cylinders"] = auto_df["cylinders"].astype(int)
auto_df["model_year"] = auto_df["model_year"].astype(int)
auto_df["origin"] = auto_df["origin"].astype(int)
build_auto(auto_df, GBDTLMRegressor(RandomForestRegressor(n_estimators = 7, max_depth = 6, random_state = 13), LinearRegression()), "GBDTLMAuto")
auto_df = load_auto("Auto")
build_chaid_auto(auto_df, "CHAIDAuto")
build_expr_auto(auto_df, "ExpressionAuto")
build_multi_auto(auto_df, EstimatorChain([("acceleration", Link(DecisionTreeRegressor(max_depth = 3, random_state = 13), augment_funcs = ["predict", "apply"]), str(True)), ("mpg", LinearRegression(), str(True))]), "MultiEstimatorChainAuto")
auto_df = load_auto("AutoNA")
build_chaid_auto(auto_df, "CHAIDAutoNA")
if "Housing" in datasets:
housing_df = load_housing("Housing")
build_housing(housing_df, GBDTLMRegressor(GradientBoostingRegressor(n_estimators = 31, random_state = 13), LinearRegression()), "GBDTLMHousing")
def build_wine(wine_df, regressor, name, eval_rows = True):
wine_X, wine_y = split_csv(wine_df)
cols = wine_X.columns.values.tolist()
memory = Memory()
memorizer = Memorizer(memory, ["subset"])
recaller = Recaller(memory, ["subset"])
regressor.controller = recaller
def make_scaler(name):
red_scaler = Alias(StandardScaler(), name = "standardScaler({}, red)".format(name))
white_scaler = Alias(StandardScaler(), name = "standardScaler({}, white)".format(name))
return SelectFirstTransformer([
("red", red_scaler, "X[0] == 'red'" if eval_rows else "X[:, 0] == 'red'"),
("white", white_scaler, "X[0] == 'white'" if eval_rows else "X[:, 0] == 'white'")
], controller = recaller, eval_rows = eval_rows)
pipeline = PMMLPipeline([
("mapper", ColumnTransformer([
("cont", Pipeline([("domain", ContinuousDomain())]), cols[0:-1]),
("cat", Pipeline([("domain", CategoricalDomain()), ("memorizer", memorizer)]), cols[-1:])
])),
("scaler", ColumnTransformer(
[(cols[idx], make_scaler(cols[idx]), [idx]) for idx in range(0, len(cols) - 1)]
)),
("regressor", regressor)
])
pipeline.fit(wine_X, wine_y)
store_pkl(pipeline, name)
quality = Series(pipeline.predict(wine_X), name = "quality")
store_csv(quality, name)
if "Wine" in datasets:
wine_df = load_wine("Wine")
def make_steps(eval_rows):
return [
("red", LinearRegression(), "X[0] == 'red'" if eval_rows else "X[:, 0] == 'red'"),
("white", LinearRegression(), "X[0] == 'white'" if eval_rows else "X[:, 0] == 'white'")
]
build_wine(wine_df, EstimatorChain(make_steps(eval_rows = True), multioutput = False), "EstimatorChainWine")
build_wine(wine_df, SelectFirstRegressor(make_steps(eval_rows = False), eval_rows = False), "SelectFirstWine", eval_rows = False)