isotree 0.1.0
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +7 -0
- data/CHANGELOG.md +3 -0
- data/LICENSE.txt +25 -0
- data/README.md +111 -0
- data/ext/isotree/ext.cpp +178 -0
- data/ext/isotree/extconf.rb +21 -0
- data/lib/isotree.rb +9 -0
- data/lib/isotree/isolation_forest.rb +94 -0
- data/lib/isotree/version.rb +3 -0
- data/vendor/isotree/LICENSE +25 -0
- data/vendor/isotree/README.md +167 -0
- data/vendor/isotree/src/Makevars +4 -0
- data/vendor/isotree/src/RcppExports.cpp +267 -0
- data/vendor/isotree/src/Rwrapper.cpp +762 -0
- data/vendor/isotree/src/crit.cpp +912 -0
- data/vendor/isotree/src/dealloc.cpp +66 -0
- data/vendor/isotree/src/dist.cpp +749 -0
- data/vendor/isotree/src/extended.cpp +790 -0
- data/vendor/isotree/src/fit_model.cpp +1068 -0
- data/vendor/isotree/src/helpers_iforest.cpp +309 -0
- data/vendor/isotree/src/impute.cpp +1205 -0
- data/vendor/isotree/src/isoforest.cpp +771 -0
- data/vendor/isotree/src/isotree.hpp +929 -0
- data/vendor/isotree/src/merge_models.cpp +116 -0
- data/vendor/isotree/src/mult.cpp +607 -0
- data/vendor/isotree/src/predict.cpp +849 -0
- data/vendor/isotree/src/serialize.cpp +262 -0
- data/vendor/isotree/src/utils.cpp +1574 -0
- metadata +154 -0
@@ -0,0 +1,762 @@
|
|
1
|
+
/* Isolation forests and variations thereof, with adjustments for incorporation
|
2
|
+
* of categorical variables and missing values.
|
3
|
+
* Writen for C++11 standard and aimed at being used in R and Python.
|
4
|
+
*
|
5
|
+
* This library is based on the following works:
|
6
|
+
* [1] Liu, Fei Tony, Kai Ming Ting, and Zhi-Hua Zhou.
|
7
|
+
* "Isolation forest."
|
8
|
+
* 2008 Eighth IEEE International Conference on Data Mining. IEEE, 2008.
|
9
|
+
* [2] Liu, Fei Tony, Kai Ming Ting, and Zhi-Hua Zhou.
|
10
|
+
* "Isolation-based anomaly detection."
|
11
|
+
* ACM Transactions on Knowledge Discovery from Data (TKDD) 6.1 (2012): 3.
|
12
|
+
* [3] Hariri, Sahand, Matias Carrasco Kind, and Robert J. Brunner.
|
13
|
+
* "Extended Isolation Forest."
|
14
|
+
* arXiv preprint arXiv:1811.02141 (2018).
|
15
|
+
* [4] Liu, Fei Tony, Kai Ming Ting, and Zhi-Hua Zhou.
|
16
|
+
* "On detecting clustered anomalies using SCiForest."
|
17
|
+
* Joint European Conference on Machine Learning and Knowledge Discovery in Databases. Springer, Berlin, Heidelberg, 2010.
|
18
|
+
* [5] https://sourceforge.net/projects/iforest/
|
19
|
+
* [6] https://math.stackexchange.com/questions/3388518/expected-number-of-paths-required-to-separate-elements-in-a-binary-tree
|
20
|
+
* [7] Quinlan, J. Ross. C4. 5: programs for machine learning. Elsevier, 2014.
|
21
|
+
* [8] Cortes, David. "Distance approximation using Isolation Forests." arXiv preprint arXiv:1910.12362 (2019).
|
22
|
+
* [9] Cortes, David. "Imputing missing values with unsupervised random trees." arXiv preprint arXiv:1911.06646 (2019).
|
23
|
+
*
|
24
|
+
* BSD 2-Clause License
|
25
|
+
* Copyright (c) 2019, David Cortes
|
26
|
+
* All rights reserved.
|
27
|
+
* Redistribution and use in source and binary forms, with or without
|
28
|
+
* modification, are permitted provided that the following conditions are met:
|
29
|
+
* * Redistributions of source code must retain the above copyright notice, this
|
30
|
+
* list of conditions and the following disclaimer.
|
31
|
+
* * Redistributions in binary form must reproduce the above copyright notice,
|
32
|
+
* this list of conditions and the following disclaimer in the documentation
|
33
|
+
* and/or other materials provided with the distribution.
|
34
|
+
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
|
35
|
+
* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
36
|
+
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
|
37
|
+
* DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
|
38
|
+
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
|
39
|
+
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
|
40
|
+
* SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
|
41
|
+
* CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
|
42
|
+
* OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
43
|
+
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
44
|
+
*/
|
45
|
+
|
46
|
+
#include <Rcpp.h>
|
47
|
+
// [[Rcpp::plugins(cpp11)]]
|
48
|
+
|
49
|
+
/* This is to serialize the model objects */
|
50
|
+
// [[Rcpp::depends(Rcereal)]]
|
51
|
+
#include <cereal/archives/binary.hpp>
|
52
|
+
#include <cereal/types/vector.hpp>
|
53
|
+
#include <sstream>
|
54
|
+
#include <string>
|
55
|
+
|
56
|
+
/* This is the package's header */
|
57
|
+
#include "isotree.hpp"
|
58
|
+
|
59
|
+
/* for model serialization and re-usage in R */
|
60
|
+
/* https://stackoverflow.com/questions/18474292/how-to-handle-c-internal-data-structure-in-r-in-order-to-allow-save-load */
|
61
|
+
/* this extra comment below the link is a workaround for Rcpp issue 675 in GitHub, do not remove it */
|
62
|
+
#include <Rinternals.h>
|
63
|
+
template <class T>
|
64
|
+
Rcpp::RawVector serialize_cpp_obj(T *model_outputs)
|
65
|
+
{
|
66
|
+
std::stringstream ss;
|
67
|
+
{
|
68
|
+
cereal::BinaryOutputArchive oarchive(ss); // Create an output archive
|
69
|
+
oarchive(*model_outputs);
|
70
|
+
}
|
71
|
+
ss.seekg(0, ss.end);
|
72
|
+
Rcpp::RawVector retval(ss.tellg());
|
73
|
+
ss.seekg(0, ss.beg);
|
74
|
+
ss.read(reinterpret_cast<char*>(&retval[0]), retval.size());
|
75
|
+
return retval;
|
76
|
+
}
|
77
|
+
|
78
|
+
// [[Rcpp::export]]
|
79
|
+
SEXP deserialize_IsoForest(Rcpp::RawVector src)
|
80
|
+
{
|
81
|
+
std::stringstream ss;
|
82
|
+
ss.write(reinterpret_cast<char*>(&src[0]), src.size());
|
83
|
+
ss.seekg(0, ss.beg);
|
84
|
+
std::unique_ptr<IsoForest> model_outputs = std::unique_ptr<IsoForest>(new IsoForest());
|
85
|
+
{
|
86
|
+
cereal::BinaryInputArchive iarchive(ss);
|
87
|
+
iarchive(*model_outputs);
|
88
|
+
}
|
89
|
+
return Rcpp::XPtr<IsoForest>(model_outputs.release(), true);
|
90
|
+
}
|
91
|
+
|
92
|
+
// [[Rcpp::export]]
|
93
|
+
SEXP deserialize_ExtIsoForest(Rcpp::RawVector src)
|
94
|
+
{
|
95
|
+
std::stringstream ss;
|
96
|
+
ss.write(reinterpret_cast<char*>(&src[0]), src.size());
|
97
|
+
ss.seekg(0, ss.beg);
|
98
|
+
std::unique_ptr<ExtIsoForest> model_outputs = std::unique_ptr<ExtIsoForest>(new ExtIsoForest());
|
99
|
+
{
|
100
|
+
cereal::BinaryInputArchive iarchive(ss);
|
101
|
+
iarchive(*model_outputs);
|
102
|
+
}
|
103
|
+
return Rcpp::XPtr<ExtIsoForest>(model_outputs.release(), true);
|
104
|
+
}
|
105
|
+
|
106
|
+
// [[Rcpp::export]]
|
107
|
+
SEXP deserialize_Imputer(Rcpp::RawVector src)
|
108
|
+
{
|
109
|
+
std::stringstream ss;
|
110
|
+
ss.write(reinterpret_cast<char*>(&src[0]), src.size());
|
111
|
+
ss.seekg(0, ss.beg);
|
112
|
+
std::unique_ptr<Imputer> imputer = std::unique_ptr<Imputer>(new Imputer());
|
113
|
+
{
|
114
|
+
cereal::BinaryInputArchive iarchive(ss);
|
115
|
+
iarchive(*imputer);
|
116
|
+
}
|
117
|
+
return Rcpp::XPtr<Imputer>(imputer.release(), true);
|
118
|
+
}
|
119
|
+
|
120
|
+
// [[Rcpp::export]]
|
121
|
+
Rcpp::LogicalVector check_null_ptr_model(SEXP ptr_model)
|
122
|
+
{
|
123
|
+
return Rcpp::LogicalVector(R_ExternalPtrAddr(ptr_model) == NULL);
|
124
|
+
}
|
125
|
+
|
126
|
+
double* set_R_nan_as_C_nan(double *x, size_t n, std::vector<double> &v, int nthreads)
|
127
|
+
{
|
128
|
+
v.assign(x, x + n);
|
129
|
+
#pragma omp parallel for schedule(static) num_threads(nthreads) shared(x, v, n)
|
130
|
+
for (size_t_for i = 0; i < n; i++)
|
131
|
+
if (isnan(v[i]))
|
132
|
+
v[i] = NAN;
|
133
|
+
return v.data();
|
134
|
+
}
|
135
|
+
|
136
|
+
double* set_R_nan_as_C_nan(double *x, size_t n, int nthreads)
|
137
|
+
{
|
138
|
+
#pragma omp parallel for schedule(static) num_threads(nthreads) shared(x, n)
|
139
|
+
for (size_t_for i = 0; i < n; i++)
|
140
|
+
if (isnan(x[i]))
|
141
|
+
x[i] = NAN;
|
142
|
+
return &x[0];
|
143
|
+
}
|
144
|
+
|
145
|
+
// [[Rcpp::export]]
|
146
|
+
Rcpp::List fit_model(Rcpp::NumericVector X_num, Rcpp::IntegerVector X_cat, Rcpp::IntegerVector ncat,
|
147
|
+
Rcpp::NumericVector Xc, Rcpp::IntegerVector Xc_ind, Rcpp::IntegerVector Xc_indptr,
|
148
|
+
Rcpp::NumericVector sample_weights, Rcpp::NumericVector col_weights,
|
149
|
+
size_t nrows, size_t ncols_numeric, size_t ncols_categ, size_t ndim, size_t ntry,
|
150
|
+
Rcpp::CharacterVector coef_type, bool coef_by_prop, bool with_replacement, bool weight_as_sample,
|
151
|
+
size_t sample_size, size_t ntrees, size_t max_depth, bool limit_depth,
|
152
|
+
bool penalize_range, bool calc_dist, bool standardize_dist, bool sq_dist,
|
153
|
+
bool calc_depth, bool standardize_depth, bool weigh_by_kurt,
|
154
|
+
double prob_pick_by_gain_avg, double prob_split_by_gain_avg,
|
155
|
+
double prob_pick_by_gain_pl, double prob_split_by_gain_pl, double min_gain,
|
156
|
+
Rcpp::CharacterVector cat_split_type, Rcpp::CharacterVector new_cat_action,
|
157
|
+
Rcpp::CharacterVector missing_action, bool all_perm,
|
158
|
+
bool build_imputer, bool output_imputations, size_t min_imp_obs,
|
159
|
+
Rcpp::CharacterVector depth_imp, Rcpp::CharacterVector weigh_imp_rows,
|
160
|
+
int random_seed, int nthreads)
|
161
|
+
{
|
162
|
+
double* numeric_data_ptr = NULL;
|
163
|
+
int* categ_data_ptr = NULL;
|
164
|
+
int* ncat_ptr = NULL;
|
165
|
+
double* Xc_ptr = NULL;
|
166
|
+
sparse_ix* Xc_ind_ptr = NULL;
|
167
|
+
sparse_ix* Xc_indptr_ptr = NULL;
|
168
|
+
double* sample_weights_ptr = NULL;
|
169
|
+
double* col_weights_ptr = NULL;
|
170
|
+
std::vector<double> Xcpp;
|
171
|
+
|
172
|
+
if (X_num.size())
|
173
|
+
{
|
174
|
+
numeric_data_ptr = &X_num[0];
|
175
|
+
if (Rcpp::as<std::string>(missing_action) != std::string("fail"))
|
176
|
+
numeric_data_ptr = set_R_nan_as_C_nan(numeric_data_ptr, nrows * ncols_numeric, Xcpp, nthreads);
|
177
|
+
}
|
178
|
+
|
179
|
+
if (X_cat.size())
|
180
|
+
{
|
181
|
+
categ_data_ptr = &X_cat[0];
|
182
|
+
ncat_ptr = &ncat[0];
|
183
|
+
}
|
184
|
+
|
185
|
+
if (Xc.size())
|
186
|
+
{
|
187
|
+
Xc_ptr = &Xc[0];
|
188
|
+
Xc_ind_ptr = &Xc_ind[0];
|
189
|
+
Xc_indptr_ptr = &Xc_indptr[0];
|
190
|
+
if (Rcpp::as<std::string>(missing_action) != std::string("fail"))
|
191
|
+
Xc_ptr = set_R_nan_as_C_nan(Xc_ptr, Xc.size(), Xcpp, nthreads);
|
192
|
+
}
|
193
|
+
|
194
|
+
if (sample_weights.size())
|
195
|
+
{
|
196
|
+
sample_weights_ptr = &sample_weights[0];
|
197
|
+
}
|
198
|
+
|
199
|
+
if (col_weights.size())
|
200
|
+
{
|
201
|
+
col_weights_ptr = &col_weights[0];
|
202
|
+
}
|
203
|
+
|
204
|
+
CoefType coef_type_C = Normal;
|
205
|
+
CategSplit cat_split_type_C = SubSet;
|
206
|
+
NewCategAction new_cat_action_C = Weighted;
|
207
|
+
MissingAction missing_action_C = Divide;
|
208
|
+
UseDepthImp depth_imp_C = Higher;
|
209
|
+
WeighImpRows weigh_imp_rows_C = Inverse;
|
210
|
+
|
211
|
+
if (Rcpp::as<std::string>(coef_type) == std::string("uniform"))
|
212
|
+
{
|
213
|
+
coef_type_C = Uniform;
|
214
|
+
}
|
215
|
+
if (Rcpp::as<std::string>(cat_split_type) == std::string("single_categ"))
|
216
|
+
{
|
217
|
+
cat_split_type_C = SingleCateg;
|
218
|
+
}
|
219
|
+
if (Rcpp::as<std::string>(new_cat_action) == std::string("smallest"))
|
220
|
+
{
|
221
|
+
new_cat_action_C = Smallest;
|
222
|
+
}
|
223
|
+
else if (Rcpp::as<std::string>(new_cat_action) == std::string("random"))
|
224
|
+
{
|
225
|
+
new_cat_action_C = Random;
|
226
|
+
}
|
227
|
+
if (Rcpp::as<std::string>(missing_action) == std::string("impute"))
|
228
|
+
{
|
229
|
+
missing_action_C = Impute;
|
230
|
+
}
|
231
|
+
else if (Rcpp::as<std::string>(missing_action) == std::string("fail"))
|
232
|
+
{
|
233
|
+
missing_action_C = Fail;
|
234
|
+
}
|
235
|
+
if (Rcpp::as<std::string>(depth_imp) == std::string("lower"))
|
236
|
+
{
|
237
|
+
depth_imp_C = Lower;
|
238
|
+
}
|
239
|
+
else if (Rcpp::as<std::string>(depth_imp) == std::string("same"))
|
240
|
+
{
|
241
|
+
depth_imp_C = Same;
|
242
|
+
}
|
243
|
+
if (Rcpp::as<std::string>(weigh_imp_rows) == std::string("prop"))
|
244
|
+
{
|
245
|
+
weigh_imp_rows_C = Prop;
|
246
|
+
}
|
247
|
+
else if (Rcpp::as<std::string>(weigh_imp_rows) == std::string("flat"))
|
248
|
+
{
|
249
|
+
weigh_imp_rows_C = Flat;
|
250
|
+
}
|
251
|
+
|
252
|
+
Rcpp::NumericVector tmat = Rcpp::NumericVector();
|
253
|
+
Rcpp::NumericMatrix dmat = Rcpp::NumericMatrix();
|
254
|
+
Rcpp::NumericVector depths = Rcpp::NumericVector();
|
255
|
+
double* tmat_ptr = NULL;
|
256
|
+
double* dmat_ptr = NULL;
|
257
|
+
double* depths_ptr = NULL;
|
258
|
+
|
259
|
+
if (calc_dist)
|
260
|
+
{
|
261
|
+
tmat = Rcpp::NumericVector((nrows * (nrows - 1)) / 2);
|
262
|
+
tmat_ptr = &tmat[0];
|
263
|
+
if (sq_dist)
|
264
|
+
{
|
265
|
+
dmat = Rcpp::NumericMatrix(nrows);
|
266
|
+
dmat_ptr = &dmat(0, 0);
|
267
|
+
}
|
268
|
+
}
|
269
|
+
|
270
|
+
if (calc_depth)
|
271
|
+
{
|
272
|
+
depths = Rcpp::NumericVector(nrows);
|
273
|
+
depths_ptr = &depths[0];
|
274
|
+
}
|
275
|
+
|
276
|
+
std::unique_ptr<IsoForest> model_ptr = std::unique_ptr<IsoForest>();
|
277
|
+
std::unique_ptr<ExtIsoForest> ext_model_ptr = std::unique_ptr<ExtIsoForest>();
|
278
|
+
std::unique_ptr<Imputer> imputer_ptr = std::unique_ptr<Imputer>();
|
279
|
+
|
280
|
+
if (ndim == 1)
|
281
|
+
model_ptr = std::unique_ptr<IsoForest>(new IsoForest());
|
282
|
+
else
|
283
|
+
ext_model_ptr = std::unique_ptr<ExtIsoForest>(new ExtIsoForest());
|
284
|
+
|
285
|
+
if (build_imputer)
|
286
|
+
imputer_ptr = std::unique_ptr<Imputer>(new Imputer());
|
287
|
+
|
288
|
+
int ret_val =
|
289
|
+
fit_iforest(model_ptr.get(), ext_model_ptr.get(),
|
290
|
+
numeric_data_ptr, ncols_numeric,
|
291
|
+
categ_data_ptr, ncols_categ, ncat_ptr,
|
292
|
+
Xc_ptr, Xc_ind_ptr, Xc_indptr_ptr,
|
293
|
+
ndim, ntry, coef_type_C, coef_by_prop,
|
294
|
+
sample_weights_ptr, with_replacement, weight_as_sample,
|
295
|
+
nrows, sample_size, ntrees, max_depth,
|
296
|
+
limit_depth, penalize_range,
|
297
|
+
standardize_dist, tmat_ptr,
|
298
|
+
depths_ptr, standardize_depth,
|
299
|
+
col_weights_ptr, weigh_by_kurt,
|
300
|
+
prob_pick_by_gain_avg, prob_split_by_gain_avg,
|
301
|
+
prob_pick_by_gain_pl, prob_split_by_gain_pl,
|
302
|
+
min_gain, missing_action_C,
|
303
|
+
cat_split_type_C, new_cat_action_C,
|
304
|
+
all_perm, imputer_ptr.get(), min_imp_obs,
|
305
|
+
depth_imp_C, weigh_imp_rows_C, output_imputations,
|
306
|
+
(uint64_t) random_seed, nthreads);
|
307
|
+
|
308
|
+
if (ret_val == EXIT_FAILURE)
|
309
|
+
{
|
310
|
+
return Rcpp::List::create(Rcpp::_["err"] = Rcpp::LogicalVector::create(1));
|
311
|
+
}
|
312
|
+
|
313
|
+
if (calc_dist && sq_dist)
|
314
|
+
tmat_to_dense(tmat_ptr, dmat_ptr, nrows, !standardize_dist);
|
315
|
+
|
316
|
+
Rcpp::RawVector serialized_obj;
|
317
|
+
if (ndim == 1)
|
318
|
+
serialized_obj = serialize_cpp_obj(model_ptr.get());
|
319
|
+
else
|
320
|
+
serialized_obj = serialize_cpp_obj(ext_model_ptr.get());
|
321
|
+
|
322
|
+
Rcpp::List outp = Rcpp::List::create(
|
323
|
+
Rcpp::_["serialized_obj"] = serialized_obj,
|
324
|
+
Rcpp::_["depths"] = depths,
|
325
|
+
Rcpp::_["tmat"] = tmat,
|
326
|
+
Rcpp::_["dmat"] = dmat
|
327
|
+
);
|
328
|
+
|
329
|
+
if (ndim == 1)
|
330
|
+
outp["model_ptr"] = Rcpp::XPtr<IsoForest>(model_ptr.release(), true);
|
331
|
+
else
|
332
|
+
outp["model_ptr"] = Rcpp::XPtr<ExtIsoForest>(ext_model_ptr.release(), true);
|
333
|
+
|
334
|
+
if (build_imputer)
|
335
|
+
{
|
336
|
+
outp["imputer_ser"] = serialize_cpp_obj(imputer_ptr.get());
|
337
|
+
outp["imputer_ptr"] = Rcpp::XPtr<Imputer>(imputer_ptr.release(), true);
|
338
|
+
}
|
339
|
+
|
340
|
+
if (output_imputations)
|
341
|
+
{
|
342
|
+
outp["imputed_num"] = Rcpp::NumericVector(Xcpp.begin(), Xcpp.end());
|
343
|
+
outp["imputed_cat"] = X_cat;
|
344
|
+
}
|
345
|
+
|
346
|
+
outp["err"] = Rcpp::LogicalVector::create(0);
|
347
|
+
|
348
|
+
return outp;
|
349
|
+
}
|
350
|
+
|
351
|
+
// [[Rcpp::export]]
|
352
|
+
Rcpp::RawVector fit_tree(SEXP model_R_ptr,
|
353
|
+
Rcpp::NumericVector X_num, Rcpp::IntegerVector X_cat, Rcpp::IntegerVector ncat,
|
354
|
+
Rcpp::NumericVector Xc, Rcpp::IntegerVector Xc_ind, Rcpp::IntegerVector Xc_indptr,
|
355
|
+
Rcpp::NumericVector sample_weights, Rcpp::NumericVector col_weights,
|
356
|
+
size_t nrows, size_t ncols_numeric, size_t ncols_categ,
|
357
|
+
size_t ndim, size_t ntry, Rcpp::CharacterVector coef_type, bool coef_by_prop,
|
358
|
+
size_t max_depth, bool limit_depth, bool penalize_range,
|
359
|
+
bool weigh_by_kurt,
|
360
|
+
double prob_pick_by_gain_avg, double prob_split_by_gain_avg,
|
361
|
+
double prob_pick_by_gain_pl, double prob_split_by_gain_pl, double min_gain,
|
362
|
+
Rcpp::CharacterVector cat_split_type, Rcpp::CharacterVector new_cat_action,
|
363
|
+
Rcpp::CharacterVector missing_action, bool build_imputer, size_t min_imp_obs, SEXP imp_R_ptr,
|
364
|
+
Rcpp::CharacterVector depth_imp, Rcpp::CharacterVector weigh_imp_rows,
|
365
|
+
bool all_perm, uint64_t random_seed)
|
366
|
+
{
|
367
|
+
double* numeric_data_ptr = NULL;
|
368
|
+
int* categ_data_ptr = NULL;
|
369
|
+
int* ncat_ptr = NULL;
|
370
|
+
double* Xc_ptr = NULL;
|
371
|
+
sparse_ix* Xc_ind_ptr = NULL;
|
372
|
+
sparse_ix* Xc_indptr_ptr = NULL;
|
373
|
+
double* sample_weights_ptr = NULL;
|
374
|
+
double* col_weights_ptr = NULL;
|
375
|
+
std::vector<double> Xcpp;
|
376
|
+
|
377
|
+
if (X_num.size())
|
378
|
+
{
|
379
|
+
numeric_data_ptr = &X_num[0];
|
380
|
+
if (Rcpp::as<std::string>(missing_action) != std::string("fail"))
|
381
|
+
numeric_data_ptr = set_R_nan_as_C_nan(numeric_data_ptr, nrows * ncols_numeric, Xcpp, 1);
|
382
|
+
}
|
383
|
+
|
384
|
+
if (X_cat.size())
|
385
|
+
{
|
386
|
+
categ_data_ptr = &X_cat[0];
|
387
|
+
ncat_ptr = &ncat[0];
|
388
|
+
}
|
389
|
+
|
390
|
+
if (Xc.size())
|
391
|
+
{
|
392
|
+
Xc_ptr = &Xc[0];
|
393
|
+
Xc_ind_ptr = &Xc_ind[0];
|
394
|
+
Xc_indptr_ptr = &Xc_indptr[0];
|
395
|
+
if (Rcpp::as<std::string>(missing_action) != std::string("fail"))
|
396
|
+
Xc_ptr = set_R_nan_as_C_nan(Xc_ptr, Xc.size(), Xcpp, 1);
|
397
|
+
}
|
398
|
+
|
399
|
+
if (sample_weights.size())
|
400
|
+
{
|
401
|
+
sample_weights_ptr = &sample_weights[0];
|
402
|
+
}
|
403
|
+
|
404
|
+
if (col_weights.size())
|
405
|
+
{
|
406
|
+
col_weights_ptr = &col_weights[0];
|
407
|
+
}
|
408
|
+
|
409
|
+
CoefType coef_type_C = Normal;
|
410
|
+
CategSplit cat_split_type_C = SubSet;
|
411
|
+
NewCategAction new_cat_action_C = Weighted;
|
412
|
+
MissingAction missing_action_C = Divide;
|
413
|
+
UseDepthImp depth_imp_C = Higher;
|
414
|
+
WeighImpRows weigh_imp_rows_C = Inverse;
|
415
|
+
|
416
|
+
if (Rcpp::as<std::string>(coef_type) == std::string("uniform"))
|
417
|
+
{
|
418
|
+
coef_type_C = Uniform;
|
419
|
+
}
|
420
|
+
if (Rcpp::as<std::string>(cat_split_type) == std::string("single_categ"))
|
421
|
+
{
|
422
|
+
cat_split_type_C = SingleCateg;
|
423
|
+
}
|
424
|
+
if (Rcpp::as<std::string>(new_cat_action) == std::string("smallest"))
|
425
|
+
{
|
426
|
+
new_cat_action_C = Smallest;
|
427
|
+
}
|
428
|
+
else if (Rcpp::as<std::string>(new_cat_action) == std::string("random"))
|
429
|
+
{
|
430
|
+
new_cat_action_C = Random;
|
431
|
+
}
|
432
|
+
if (Rcpp::as<std::string>(missing_action) == std::string("impute"))
|
433
|
+
{
|
434
|
+
missing_action_C = Impute;
|
435
|
+
}
|
436
|
+
else if (Rcpp::as<std::string>(missing_action) == std::string("fail"))
|
437
|
+
{
|
438
|
+
missing_action_C = Fail;
|
439
|
+
}
|
440
|
+
if (Rcpp::as<std::string>(depth_imp) == std::string("lower"))
|
441
|
+
{
|
442
|
+
depth_imp_C = Lower;
|
443
|
+
}
|
444
|
+
else if (Rcpp::as<std::string>(depth_imp) == std::string("same"))
|
445
|
+
{
|
446
|
+
depth_imp_C = Same;
|
447
|
+
}
|
448
|
+
if (Rcpp::as<std::string>(weigh_imp_rows) == std::string("prop"))
|
449
|
+
{
|
450
|
+
weigh_imp_rows_C = Prop;
|
451
|
+
}
|
452
|
+
else if (Rcpp::as<std::string>(weigh_imp_rows) == std::string("flat"))
|
453
|
+
{
|
454
|
+
weigh_imp_rows_C = Flat;
|
455
|
+
}
|
456
|
+
|
457
|
+
IsoForest* model_ptr = NULL;
|
458
|
+
ExtIsoForest* ext_model_ptr = NULL;
|
459
|
+
Imputer* imputer_ptr = NULL;
|
460
|
+
if (ndim == 1)
|
461
|
+
model_ptr = static_cast<IsoForest*>(R_ExternalPtrAddr(model_R_ptr));
|
462
|
+
else
|
463
|
+
ext_model_ptr = static_cast<ExtIsoForest*>(R_ExternalPtrAddr(model_R_ptr));
|
464
|
+
|
465
|
+
std::vector<ImputeNode> *imp_ptr = NULL;
|
466
|
+
if (build_imputer)
|
467
|
+
{
|
468
|
+
imputer_ptr = static_cast<Imputer*>(R_ExternalPtrAddr(imp_R_ptr));
|
469
|
+
imputer_ptr->imputer_tree.emplace_back();
|
470
|
+
imp_ptr = &imputer_ptr->imputer_tree.back();
|
471
|
+
}
|
472
|
+
|
473
|
+
add_tree(model_ptr, ext_model_ptr,
|
474
|
+
numeric_data_ptr, ncols_numeric,
|
475
|
+
categ_data_ptr, ncols_categ, ncat_ptr,
|
476
|
+
Xc_ptr, Xc_ind_ptr, Xc_indptr_ptr,
|
477
|
+
ndim, ntry, coef_type_C, coef_by_prop,
|
478
|
+
sample_weights_ptr,
|
479
|
+
nrows, max_depth,
|
480
|
+
limit_depth, penalize_range,
|
481
|
+
col_weights_ptr, weigh_by_kurt,
|
482
|
+
prob_pick_by_gain_avg, prob_split_by_gain_avg,
|
483
|
+
prob_pick_by_gain_pl, prob_split_by_gain_pl,
|
484
|
+
min_gain, missing_action_C,
|
485
|
+
cat_split_type_C, new_cat_action_C,
|
486
|
+
depth_imp_C, weigh_imp_rows_C, all_perm,
|
487
|
+
imp_ptr, min_imp_obs, (uint64_t)random_seed);
|
488
|
+
|
489
|
+
if (ndim == 1)
|
490
|
+
return serialize_cpp_obj(model_ptr);
|
491
|
+
else
|
492
|
+
return serialize_cpp_obj(ext_model_ptr);
|
493
|
+
}
|
494
|
+
|
495
|
+
// [[Rcpp::export]]
|
496
|
+
void predict_iso(SEXP model_R_ptr, Rcpp::NumericVector outp, Rcpp::IntegerVector tree_num, bool is_extended,
|
497
|
+
Rcpp::NumericVector X_num, Rcpp::IntegerVector X_cat,
|
498
|
+
Rcpp::NumericVector Xc, Rcpp::IntegerVector Xc_ind, Rcpp::IntegerVector Xc_indptr,
|
499
|
+
Rcpp::NumericVector Xr, Rcpp::IntegerVector Xr_ind, Rcpp::IntegerVector Xr_indptr,
|
500
|
+
size_t nrows, int nthreads, bool standardize)
|
501
|
+
{
|
502
|
+
double* numeric_data_ptr = NULL;
|
503
|
+
int* categ_data_ptr = NULL;
|
504
|
+
double* Xc_ptr = NULL;
|
505
|
+
sparse_ix* Xc_ind_ptr = NULL;
|
506
|
+
sparse_ix* Xc_indptr_ptr = NULL;
|
507
|
+
double* Xr_ptr = NULL;
|
508
|
+
sparse_ix* Xr_ind_ptr = NULL;
|
509
|
+
sparse_ix* Xr_indptr_ptr = NULL;
|
510
|
+
sparse_ix* tree_num_ptr = NULL;
|
511
|
+
std::vector<double> Xcpp;
|
512
|
+
|
513
|
+
if (X_num.size())
|
514
|
+
{
|
515
|
+
numeric_data_ptr = &X_num[0];
|
516
|
+
}
|
517
|
+
|
518
|
+
if (X_cat.size())
|
519
|
+
{
|
520
|
+
categ_data_ptr = &X_cat[0];
|
521
|
+
}
|
522
|
+
|
523
|
+
if (Xc.size())
|
524
|
+
{
|
525
|
+
Xc_ptr = &Xc[0];
|
526
|
+
Xc_ind_ptr = &Xc_ind[0];
|
527
|
+
Xc_indptr_ptr = &Xc_indptr[0];
|
528
|
+
}
|
529
|
+
|
530
|
+
if (Xr.size())
|
531
|
+
{
|
532
|
+
Xr_ptr = &Xr[0];
|
533
|
+
Xr_ind_ptr = &Xr_ind[0];
|
534
|
+
Xr_indptr_ptr = &Xr_indptr[0];
|
535
|
+
}
|
536
|
+
|
537
|
+
if (tree_num.size())
|
538
|
+
{
|
539
|
+
tree_num_ptr = &tree_num[0];
|
540
|
+
}
|
541
|
+
|
542
|
+
double* depths_ptr = &outp[0];
|
543
|
+
|
544
|
+
IsoForest* model_ptr = NULL;
|
545
|
+
ExtIsoForest* ext_model_ptr = NULL;
|
546
|
+
if (is_extended)
|
547
|
+
ext_model_ptr = static_cast<ExtIsoForest*>(R_ExternalPtrAddr(model_R_ptr));
|
548
|
+
else
|
549
|
+
model_ptr = static_cast<IsoForest*>(R_ExternalPtrAddr(model_R_ptr));
|
550
|
+
|
551
|
+
MissingAction missing_action = is_extended?
|
552
|
+
ext_model_ptr->missing_action
|
553
|
+
:
|
554
|
+
model_ptr->missing_action;
|
555
|
+
if (missing_action != Fail)
|
556
|
+
{
|
557
|
+
if (X_num.size()) numeric_data_ptr = set_R_nan_as_C_nan(numeric_data_ptr, X_num.size(), Xcpp, nthreads);
|
558
|
+
if (Xc.size()) Xc_ptr = set_R_nan_as_C_nan(Xc_ptr, Xc.size(), Xcpp, nthreads);
|
559
|
+
if (Xr.size()) Xr_ptr = set_R_nan_as_C_nan(Xr_ptr, Xr.size(), Xcpp, nthreads);
|
560
|
+
}
|
561
|
+
|
562
|
+
predict_iforest(numeric_data_ptr, categ_data_ptr,
|
563
|
+
Xc_ptr, Xc_ind_ptr, Xc_indptr_ptr,
|
564
|
+
Xr_ptr, Xr_ind_ptr, Xr_indptr_ptr,
|
565
|
+
nrows, nthreads, standardize,
|
566
|
+
model_ptr, ext_model_ptr,
|
567
|
+
depths_ptr, tree_num_ptr);
|
568
|
+
}
|
569
|
+
|
570
|
+
// [[Rcpp::export]]
|
571
|
+
void dist_iso(SEXP model_R_ptr, Rcpp::NumericVector tmat, Rcpp::NumericVector dmat,
|
572
|
+
Rcpp::NumericVector rmat, bool is_extended,
|
573
|
+
Rcpp::NumericVector X_num, Rcpp::IntegerVector X_cat,
|
574
|
+
Rcpp::NumericVector Xc, Rcpp::IntegerVector Xc_ind, Rcpp::IntegerVector Xc_indptr,
|
575
|
+
size_t nrows, int nthreads, bool assume_full_distr,
|
576
|
+
bool standardize_dist, bool sq_dist, size_t n_from)
|
577
|
+
{
|
578
|
+
double* numeric_data_ptr = NULL;
|
579
|
+
int* categ_data_ptr = NULL;
|
580
|
+
double* Xc_ptr = NULL;
|
581
|
+
sparse_ix* Xc_ind_ptr = NULL;
|
582
|
+
sparse_ix* Xc_indptr_ptr = NULL;
|
583
|
+
std::vector<double> Xcpp;
|
584
|
+
|
585
|
+
if (X_num.size())
|
586
|
+
{
|
587
|
+
numeric_data_ptr = &X_num[0];
|
588
|
+
}
|
589
|
+
|
590
|
+
if (X_cat.size())
|
591
|
+
{
|
592
|
+
categ_data_ptr = &X_cat[0];
|
593
|
+
}
|
594
|
+
|
595
|
+
if (Xc.size())
|
596
|
+
{
|
597
|
+
Xc_ptr = &Xc[0];
|
598
|
+
Xc_ind_ptr = &Xc_ind[0];
|
599
|
+
Xc_indptr_ptr = &Xc_indptr[0];
|
600
|
+
}
|
601
|
+
|
602
|
+
double* tmat_ptr = n_from? (double*)NULL : &tmat[0];
|
603
|
+
double* dmat_ptr = (sq_dist & !n_from)? &dmat[0] : NULL;
|
604
|
+
double* rmat_ptr = n_from? &rmat[0] : NULL;
|
605
|
+
|
606
|
+
IsoForest* model_ptr = NULL;
|
607
|
+
ExtIsoForest* ext_model_ptr = NULL;
|
608
|
+
if (is_extended)
|
609
|
+
ext_model_ptr = static_cast<ExtIsoForest*>(R_ExternalPtrAddr(model_R_ptr));
|
610
|
+
else
|
611
|
+
model_ptr = static_cast<IsoForest*>(R_ExternalPtrAddr(model_R_ptr));
|
612
|
+
|
613
|
+
|
614
|
+
MissingAction missing_action = is_extended?
|
615
|
+
ext_model_ptr->missing_action
|
616
|
+
:
|
617
|
+
model_ptr->missing_action;
|
618
|
+
if (missing_action != Fail)
|
619
|
+
{
|
620
|
+
if (X_num.size()) numeric_data_ptr = set_R_nan_as_C_nan(numeric_data_ptr, X_num.size(), Xcpp, nthreads);
|
621
|
+
if (Xc.size()) Xc_ptr = set_R_nan_as_C_nan(Xc_ptr, Xc.size(), Xcpp, nthreads);
|
622
|
+
}
|
623
|
+
|
624
|
+
|
625
|
+
calc_similarity(numeric_data_ptr, categ_data_ptr,
|
626
|
+
Xc_ptr, Xc_ind_ptr, Xc_indptr_ptr,
|
627
|
+
nrows, nthreads, assume_full_distr, standardize_dist,
|
628
|
+
model_ptr, ext_model_ptr,
|
629
|
+
tmat_ptr, rmat_ptr, n_from);
|
630
|
+
|
631
|
+
if (sq_dist & !n_from)
|
632
|
+
tmat_to_dense(tmat_ptr, dmat_ptr, nrows, !standardize_dist);
|
633
|
+
}
|
634
|
+
|
635
|
+
// [[Rcpp::export]]
|
636
|
+
Rcpp::List impute_iso(SEXP model_R_ptr, SEXP imputer_R_ptr, bool is_extended,
|
637
|
+
Rcpp::NumericVector X_num, Rcpp::IntegerVector X_cat,
|
638
|
+
Rcpp::NumericVector Xr, Rcpp::IntegerVector Xr_ind, Rcpp::IntegerVector Xr_indptr,
|
639
|
+
size_t nrows, int nthreads)
|
640
|
+
{
|
641
|
+
double* numeric_data_ptr = NULL;
|
642
|
+
int* categ_data_ptr = NULL;
|
643
|
+
double* Xr_ptr = NULL;
|
644
|
+
sparse_ix* Xr_ind_ptr = NULL;
|
645
|
+
sparse_ix* Xr_indptr_ptr = NULL;
|
646
|
+
|
647
|
+
if (X_num.size())
|
648
|
+
{
|
649
|
+
numeric_data_ptr = &X_num[0];
|
650
|
+
}
|
651
|
+
|
652
|
+
if (X_cat.size())
|
653
|
+
{
|
654
|
+
categ_data_ptr = &X_cat[0];
|
655
|
+
}
|
656
|
+
|
657
|
+
if (Xr.size())
|
658
|
+
{
|
659
|
+
Xr_ptr = &Xr[0];
|
660
|
+
Xr_ind_ptr = &Xr_ind[0];
|
661
|
+
Xr_indptr_ptr = &Xr_indptr[0];
|
662
|
+
}
|
663
|
+
|
664
|
+
if (X_num.size()) numeric_data_ptr = set_R_nan_as_C_nan(numeric_data_ptr, X_num.size(), nthreads);
|
665
|
+
if (Xr.size()) Xr_ptr = set_R_nan_as_C_nan(Xr_ptr, Xr.size(), nthreads);
|
666
|
+
|
667
|
+
IsoForest* model_ptr = NULL;
|
668
|
+
ExtIsoForest* ext_model_ptr = NULL;
|
669
|
+
if (is_extended)
|
670
|
+
ext_model_ptr = static_cast<ExtIsoForest*>(R_ExternalPtrAddr(model_R_ptr));
|
671
|
+
else
|
672
|
+
model_ptr = static_cast<IsoForest*>(R_ExternalPtrAddr(model_R_ptr));
|
673
|
+
|
674
|
+
Imputer* imputer_ptr = static_cast<Imputer*>(R_ExternalPtrAddr(imputer_R_ptr));
|
675
|
+
|
676
|
+
|
677
|
+
impute_missing_values(numeric_data_ptr, categ_data_ptr,
|
678
|
+
Xr_ptr, Xr_ind_ptr, Xr_indptr_ptr,
|
679
|
+
nrows, nthreads,
|
680
|
+
model_ptr, ext_model_ptr,
|
681
|
+
*imputer_ptr);
|
682
|
+
|
683
|
+
return Rcpp::List::create(
|
684
|
+
Rcpp::_["X_num"] = Xr.size()? Xr : X_num,
|
685
|
+
Rcpp::_["X_cat"] = X_cat
|
686
|
+
);
|
687
|
+
}
|
688
|
+
|
689
|
+
// [[Rcpp::export]]
|
690
|
+
Rcpp::List get_n_nodes(SEXP model_R_ptr, bool is_extended, int nthreads)
|
691
|
+
{
|
692
|
+
size_t ntrees;
|
693
|
+
IsoForest* model_ptr = NULL;
|
694
|
+
ExtIsoForest* ext_model_ptr = NULL;
|
695
|
+
if (is_extended)
|
696
|
+
{
|
697
|
+
ext_model_ptr = static_cast<ExtIsoForest*>(R_ExternalPtrAddr(model_R_ptr));
|
698
|
+
ntrees = ext_model_ptr->hplanes.size();
|
699
|
+
}
|
700
|
+
else
|
701
|
+
{
|
702
|
+
model_ptr = static_cast<IsoForest*>(R_ExternalPtrAddr(model_R_ptr));
|
703
|
+
ntrees = model_ptr->trees.size();
|
704
|
+
}
|
705
|
+
|
706
|
+
Rcpp::IntegerVector n_nodes(ntrees);
|
707
|
+
Rcpp::IntegerVector n_terminal(ntrees);
|
708
|
+
if (is_extended)
|
709
|
+
get_num_nodes(*ext_model_ptr, &n_nodes[0], &n_terminal[0], nthreads);
|
710
|
+
else
|
711
|
+
get_num_nodes(*model_ptr, &n_nodes[0], &n_terminal[0], nthreads);
|
712
|
+
|
713
|
+
return Rcpp::List::create(
|
714
|
+
Rcpp::_["total"] = n_nodes,
|
715
|
+
Rcpp::_["terminal"] = n_terminal
|
716
|
+
);
|
717
|
+
}
|
718
|
+
|
719
|
+
// [[Rcpp::export]]
|
720
|
+
Rcpp::List append_trees_from_other(SEXP model_R_ptr, SEXP other_R_ptr,
|
721
|
+
SEXP imp_R_ptr, SEXP oimp_R_ptr,
|
722
|
+
bool is_extended)
|
723
|
+
{
|
724
|
+
Rcpp::List out;
|
725
|
+
IsoForest* model_ptr = NULL;
|
726
|
+
IsoForest* other_ptr = NULL;
|
727
|
+
ExtIsoForest* ext_model_ptr = NULL;
|
728
|
+
ExtIsoForest* ext_other_ptr = NULL;
|
729
|
+
Imputer* imputer_ptr = NULL;
|
730
|
+
Imputer* oimputer_ptr = NULL;
|
731
|
+
|
732
|
+
if (is_extended) {
|
733
|
+
ext_model_ptr = static_cast<ExtIsoForest*>(R_ExternalPtrAddr(model_R_ptr));
|
734
|
+
ext_other_ptr = static_cast<ExtIsoForest*>(R_ExternalPtrAddr(other_R_ptr));
|
735
|
+
} else {
|
736
|
+
model_ptr = static_cast<IsoForest*>(R_ExternalPtrAddr(model_R_ptr));
|
737
|
+
other_ptr = static_cast<IsoForest*>(R_ExternalPtrAddr(other_R_ptr));
|
738
|
+
}
|
739
|
+
|
740
|
+
if (!Rf_isNull(imp_R_ptr) && !Rf_isNull(oimp_R_ptr) &&
|
741
|
+
R_ExternalPtrAddr(imp_R_ptr) != NULL &&
|
742
|
+
R_ExternalPtrAddr(oimp_R_ptr) != NULL)
|
743
|
+
{
|
744
|
+
imputer_ptr = static_cast<Imputer*>(R_ExternalPtrAddr(imp_R_ptr));
|
745
|
+
oimputer_ptr = static_cast<Imputer*>(R_ExternalPtrAddr(oimp_R_ptr));
|
746
|
+
}
|
747
|
+
|
748
|
+
merge_models(model_ptr, other_ptr,
|
749
|
+
ext_model_ptr, ext_other_ptr,
|
750
|
+
imputer_ptr, oimputer_ptr);
|
751
|
+
|
752
|
+
|
753
|
+
if (is_extended)
|
754
|
+
out["serialized"] = serialize_cpp_obj(ext_model_ptr);
|
755
|
+
else
|
756
|
+
out["serialized"] = serialize_cpp_obj(model_ptr);
|
757
|
+
|
758
|
+
if (imputer_ptr != NULL && oimputer_ptr != NULL)
|
759
|
+
out["imp_ser"] = serialize_cpp_obj(imputer_ptr);
|
760
|
+
|
761
|
+
return out;
|
762
|
+
}
|