isotree 0.1.1 → 0.2.0
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +4 -4
- data/CHANGELOG.md +25 -0
- data/LICENSE.txt +2 -1
- data/README.md +57 -6
- data/ext/isotree/ext.cpp +170 -39
- data/ext/isotree/extconf.rb +3 -3
- data/lib/isotree.rb +2 -0
- data/lib/isotree/dataset.rb +73 -0
- data/lib/isotree/isolation_forest.rb +182 -29
- data/lib/isotree/version.rb +1 -1
- data/vendor/cereal/LICENSE +24 -0
- data/vendor/cereal/README.md +85 -0
- data/vendor/cereal/include/cereal/access.hpp +351 -0
- data/vendor/cereal/include/cereal/archives/adapters.hpp +163 -0
- data/vendor/cereal/include/cereal/archives/binary.hpp +169 -0
- data/vendor/cereal/include/cereal/archives/json.hpp +1019 -0
- data/vendor/cereal/include/cereal/archives/portable_binary.hpp +334 -0
- data/vendor/cereal/include/cereal/archives/xml.hpp +956 -0
- data/vendor/cereal/include/cereal/cereal.hpp +1089 -0
- data/vendor/cereal/include/cereal/details/helpers.hpp +422 -0
- data/vendor/cereal/include/cereal/details/polymorphic_impl.hpp +796 -0
- data/vendor/cereal/include/cereal/details/polymorphic_impl_fwd.hpp +65 -0
- data/vendor/cereal/include/cereal/details/static_object.hpp +127 -0
- data/vendor/cereal/include/cereal/details/traits.hpp +1411 -0
- data/vendor/cereal/include/cereal/details/util.hpp +84 -0
- data/vendor/cereal/include/cereal/external/base64.hpp +134 -0
- data/vendor/cereal/include/cereal/external/rapidjson/allocators.h +284 -0
- data/vendor/cereal/include/cereal/external/rapidjson/cursorstreamwrapper.h +78 -0
- data/vendor/cereal/include/cereal/external/rapidjson/document.h +2652 -0
- data/vendor/cereal/include/cereal/external/rapidjson/encodedstream.h +299 -0
- data/vendor/cereal/include/cereal/external/rapidjson/encodings.h +716 -0
- data/vendor/cereal/include/cereal/external/rapidjson/error/en.h +74 -0
- data/vendor/cereal/include/cereal/external/rapidjson/error/error.h +161 -0
- data/vendor/cereal/include/cereal/external/rapidjson/filereadstream.h +99 -0
- data/vendor/cereal/include/cereal/external/rapidjson/filewritestream.h +104 -0
- data/vendor/cereal/include/cereal/external/rapidjson/fwd.h +151 -0
- data/vendor/cereal/include/cereal/external/rapidjson/internal/biginteger.h +290 -0
- data/vendor/cereal/include/cereal/external/rapidjson/internal/diyfp.h +271 -0
- data/vendor/cereal/include/cereal/external/rapidjson/internal/dtoa.h +245 -0
- data/vendor/cereal/include/cereal/external/rapidjson/internal/ieee754.h +78 -0
- data/vendor/cereal/include/cereal/external/rapidjson/internal/itoa.h +308 -0
- data/vendor/cereal/include/cereal/external/rapidjson/internal/meta.h +186 -0
- data/vendor/cereal/include/cereal/external/rapidjson/internal/pow10.h +55 -0
- data/vendor/cereal/include/cereal/external/rapidjson/internal/regex.h +740 -0
- data/vendor/cereal/include/cereal/external/rapidjson/internal/stack.h +232 -0
- data/vendor/cereal/include/cereal/external/rapidjson/internal/strfunc.h +69 -0
- data/vendor/cereal/include/cereal/external/rapidjson/internal/strtod.h +290 -0
- data/vendor/cereal/include/cereal/external/rapidjson/internal/swap.h +46 -0
- data/vendor/cereal/include/cereal/external/rapidjson/istreamwrapper.h +128 -0
- data/vendor/cereal/include/cereal/external/rapidjson/memorybuffer.h +70 -0
- data/vendor/cereal/include/cereal/external/rapidjson/memorystream.h +71 -0
- data/vendor/cereal/include/cereal/external/rapidjson/msinttypes/inttypes.h +316 -0
- data/vendor/cereal/include/cereal/external/rapidjson/msinttypes/stdint.h +300 -0
- data/vendor/cereal/include/cereal/external/rapidjson/ostreamwrapper.h +81 -0
- data/vendor/cereal/include/cereal/external/rapidjson/pointer.h +1414 -0
- data/vendor/cereal/include/cereal/external/rapidjson/prettywriter.h +277 -0
- data/vendor/cereal/include/cereal/external/rapidjson/rapidjson.h +656 -0
- data/vendor/cereal/include/cereal/external/rapidjson/reader.h +2230 -0
- data/vendor/cereal/include/cereal/external/rapidjson/schema.h +2497 -0
- data/vendor/cereal/include/cereal/external/rapidjson/stream.h +223 -0
- data/vendor/cereal/include/cereal/external/rapidjson/stringbuffer.h +121 -0
- data/vendor/cereal/include/cereal/external/rapidjson/writer.h +709 -0
- data/vendor/cereal/include/cereal/external/rapidxml/license.txt +52 -0
- data/vendor/cereal/include/cereal/external/rapidxml/manual.html +406 -0
- data/vendor/cereal/include/cereal/external/rapidxml/rapidxml.hpp +2624 -0
- data/vendor/cereal/include/cereal/external/rapidxml/rapidxml_iterators.hpp +175 -0
- data/vendor/cereal/include/cereal/external/rapidxml/rapidxml_print.hpp +428 -0
- data/vendor/cereal/include/cereal/external/rapidxml/rapidxml_utils.hpp +123 -0
- data/vendor/cereal/include/cereal/macros.hpp +154 -0
- data/vendor/cereal/include/cereal/specialize.hpp +139 -0
- data/vendor/cereal/include/cereal/types/array.hpp +79 -0
- data/vendor/cereal/include/cereal/types/atomic.hpp +55 -0
- data/vendor/cereal/include/cereal/types/base_class.hpp +203 -0
- data/vendor/cereal/include/cereal/types/bitset.hpp +176 -0
- data/vendor/cereal/include/cereal/types/boost_variant.hpp +164 -0
- data/vendor/cereal/include/cereal/types/chrono.hpp +72 -0
- data/vendor/cereal/include/cereal/types/common.hpp +129 -0
- data/vendor/cereal/include/cereal/types/complex.hpp +56 -0
- data/vendor/cereal/include/cereal/types/concepts/pair_associative_container.hpp +73 -0
- data/vendor/cereal/include/cereal/types/deque.hpp +62 -0
- data/vendor/cereal/include/cereal/types/forward_list.hpp +68 -0
- data/vendor/cereal/include/cereal/types/functional.hpp +43 -0
- data/vendor/cereal/include/cereal/types/list.hpp +62 -0
- data/vendor/cereal/include/cereal/types/map.hpp +36 -0
- data/vendor/cereal/include/cereal/types/memory.hpp +425 -0
- data/vendor/cereal/include/cereal/types/optional.hpp +66 -0
- data/vendor/cereal/include/cereal/types/polymorphic.hpp +483 -0
- data/vendor/cereal/include/cereal/types/queue.hpp +132 -0
- data/vendor/cereal/include/cereal/types/set.hpp +103 -0
- data/vendor/cereal/include/cereal/types/stack.hpp +76 -0
- data/vendor/cereal/include/cereal/types/string.hpp +61 -0
- data/vendor/cereal/include/cereal/types/tuple.hpp +123 -0
- data/vendor/cereal/include/cereal/types/unordered_map.hpp +36 -0
- data/vendor/cereal/include/cereal/types/unordered_set.hpp +99 -0
- data/vendor/cereal/include/cereal/types/utility.hpp +47 -0
- data/vendor/cereal/include/cereal/types/valarray.hpp +89 -0
- data/vendor/cereal/include/cereal/types/variant.hpp +109 -0
- data/vendor/cereal/include/cereal/types/vector.hpp +112 -0
- data/vendor/cereal/include/cereal/version.hpp +52 -0
- data/vendor/isotree/LICENSE +1 -1
- data/vendor/isotree/README.md +7 -2
- data/vendor/isotree/src/RcppExports.cpp +44 -4
- data/vendor/isotree/src/Rwrapper.cpp +141 -51
- data/vendor/isotree/src/crit.cpp +1 -1
- data/vendor/isotree/src/dealloc.cpp +1 -1
- data/vendor/isotree/src/dist.cpp +6 -6
- data/vendor/isotree/src/extended.cpp +5 -5
- data/vendor/isotree/src/fit_model.cpp +27 -5
- data/vendor/isotree/src/helpers_iforest.cpp +26 -11
- data/vendor/isotree/src/impute.cpp +7 -7
- data/vendor/isotree/src/isoforest.cpp +7 -7
- data/vendor/isotree/src/isotree.hpp +27 -5
- data/vendor/isotree/src/merge_models.cpp +1 -1
- data/vendor/isotree/src/mult.cpp +1 -1
- data/vendor/isotree/src/predict.cpp +20 -16
- data/vendor/isotree/src/serialize.cpp +1 -1
- data/vendor/isotree/src/sql.cpp +545 -0
- data/vendor/isotree/src/utils.cpp +36 -44
- metadata +102 -81
@@ -22,7 +22,7 @@
|
|
22
22
|
* [9] Cortes, David. "Imputing missing values with unsupervised random trees." arXiv preprint arXiv:1911.06646 (2019).
|
23
23
|
*
|
24
24
|
* BSD 2-Clause License
|
25
|
-
* Copyright (c)
|
25
|
+
* Copyright (c) 2020, David Cortes
|
26
26
|
* All rights reserved.
|
27
27
|
* Redistribution and use in source and binary forms, with or without
|
28
28
|
* modification, are permitted provided that the following conditions are met:
|
data/vendor/isotree/src/mult.cpp
CHANGED
@@ -22,7 +22,7 @@
|
|
22
22
|
* [9] Cortes, David. "Imputing missing values with unsupervised random trees." arXiv preprint arXiv:1911.06646 (2019).
|
23
23
|
*
|
24
24
|
* BSD 2-Clause License
|
25
|
-
* Copyright (c)
|
25
|
+
* Copyright (c) 2020, David Cortes
|
26
26
|
* All rights reserved.
|
27
27
|
* Redistribution and use in source and binary forms, with or without
|
28
28
|
* modification, are permitted provided that the following conditions are met:
|
@@ -22,7 +22,7 @@
|
|
22
22
|
* [9] Cortes, David. "Imputing missing values with unsupervised random trees." arXiv preprint arXiv:1911.06646 (2019).
|
23
23
|
*
|
24
24
|
* BSD 2-Clause License
|
25
|
-
* Copyright (c)
|
25
|
+
* Copyright (c) 2020, David Cortes
|
26
26
|
* All rights reserved.
|
27
27
|
* Redistribution and use in source and binary forms, with or without
|
28
28
|
* modification, are permitted provided that the following conditions are met:
|
@@ -137,7 +137,7 @@ void predict_iforest(double numeric_data[], int categ_data[],
|
|
137
137
|
if (
|
138
138
|
model_outputs->missing_action == Fail &&
|
139
139
|
(model_outputs->new_cat_action != Weighted || prediction_data.categ_data == NULL) &&
|
140
|
-
prediction_data.
|
140
|
+
prediction_data.Xc_indptr == NULL && prediction_data.Xr_indptr == NULL
|
141
141
|
)
|
142
142
|
{
|
143
143
|
#pragma omp parallel for schedule(static) num_threads(nthreads) shared(nrows, model_outputs, prediction_data, output_depths, tree_num)
|
@@ -180,8 +180,8 @@ void predict_iforest(double numeric_data[], int categ_data[],
|
|
180
180
|
if (
|
181
181
|
model_outputs_ext->missing_action == Fail &&
|
182
182
|
prediction_data.categ_data == NULL &&
|
183
|
-
prediction_data.
|
184
|
-
prediction_data.
|
183
|
+
prediction_data.Xc_indptr == NULL &&
|
184
|
+
prediction_data.Xr_indptr == NULL
|
185
185
|
)
|
186
186
|
{
|
187
187
|
#pragma omp parallel for schedule(static) num_threads(nthreads) shared(nrows, model_outputs_ext, prediction_data, output_depths, tree_num)
|
@@ -252,6 +252,8 @@ void predict_iforest(double numeric_data[], int categ_data[],
|
|
252
252
|
}
|
253
253
|
|
254
254
|
|
255
|
+
/* TODO: these functions would be faster if done with row-major order,
|
256
|
+
should at least give the option of taking arrays as row-major. */
|
255
257
|
void traverse_itree_no_recurse(std::vector<IsoTree> &tree,
|
256
258
|
IsoForest &model_outputs,
|
257
259
|
PredictionData &prediction_data,
|
@@ -381,7 +383,7 @@ double traverse_itree(std::vector<IsoTree> &tree,
|
|
381
383
|
double range_penalty = 0;
|
382
384
|
|
383
385
|
sparse_ix *row_st = NULL, *row_end = NULL;
|
384
|
-
if (prediction_data.
|
386
|
+
if (prediction_data.Xr_indptr != NULL)
|
385
387
|
{
|
386
388
|
row_st = prediction_data.Xr_ind + prediction_data.Xr_indptr[row];
|
387
389
|
row_end = prediction_data.Xr_ind + prediction_data.Xr_indptr[row + 1];
|
@@ -406,12 +408,12 @@ double traverse_itree(std::vector<IsoTree> &tree,
|
|
406
408
|
case Numeric:
|
407
409
|
{
|
408
410
|
|
409
|
-
if (prediction_data.
|
411
|
+
if (prediction_data.Xc_indptr == NULL && prediction_data.Xr_indptr == NULL)
|
410
412
|
xval = prediction_data.numeric_data[row + tree[curr_lev].col_num * prediction_data.nrows];
|
411
|
-
else if (
|
412
|
-
xval = extract_spR(prediction_data, row_st, row_end, tree[curr_lev].col_num);
|
413
|
-
else
|
413
|
+
else if (prediction_data.Xc_indptr != NULL)
|
414
414
|
xval = extract_spC(prediction_data, row, tree[curr_lev].col_num);
|
415
|
+
else
|
416
|
+
xval = extract_spR(prediction_data, row_st, row_end, tree[curr_lev].col_num);
|
415
417
|
|
416
418
|
if (isnan(xval))
|
417
419
|
{
|
@@ -682,7 +684,7 @@ void traverse_hplane(std::vector<IsoHPlane> &hplane,
|
|
682
684
|
size_t ncols_numeric, ncols_categ;
|
683
685
|
|
684
686
|
sparse_ix *row_st = NULL, *row_end = NULL;
|
685
|
-
if (prediction_data.
|
687
|
+
if (prediction_data.Xr_indptr != NULL)
|
686
688
|
{
|
687
689
|
row_st = prediction_data.Xr_ind + prediction_data.Xr_indptr[row];
|
688
690
|
row_end = prediction_data.Xr_ind + prediction_data.Xr_indptr[row + 1];
|
@@ -712,12 +714,12 @@ void traverse_hplane(std::vector<IsoHPlane> &hplane,
|
|
712
714
|
{
|
713
715
|
case Numeric:
|
714
716
|
{
|
715
|
-
if (prediction_data.
|
717
|
+
if (prediction_data.Xc_indptr == NULL && prediction_data.Xr_indptr == NULL)
|
716
718
|
xval = prediction_data.numeric_data[row + hplane[curr_lev].col_num[col] * prediction_data.nrows];
|
717
|
-
else if (
|
718
|
-
xval = extract_spR(prediction_data, row_st, row_end, hplane[curr_lev].col_num[col]);
|
719
|
-
else
|
719
|
+
else if (prediction_data.Xc_indptr != NULL)
|
720
720
|
xval = extract_spC(prediction_data, row, hplane[curr_lev].col_num[col]);
|
721
|
+
else
|
722
|
+
xval = extract_spR(prediction_data, row_st, row_end, hplane[curr_lev].col_num[col]);
|
721
723
|
|
722
724
|
if (is_na_or_inf(xval))
|
723
725
|
{
|
@@ -805,16 +807,18 @@ double extract_spC(PredictionData &prediction_data, size_t row, size_t col_num)
|
|
805
807
|
||
|
806
808
|
*search_res != row
|
807
809
|
)
|
808
|
-
return 0
|
810
|
+
return 0.;
|
809
811
|
else
|
810
812
|
return prediction_data.Xc[search_res - prediction_data.Xc_ind];
|
811
813
|
}
|
812
814
|
|
813
815
|
double extract_spR(PredictionData &prediction_data, sparse_ix *row_st, sparse_ix *row_end, size_t col_num)
|
814
816
|
{
|
817
|
+
if (row_end == row_st)
|
818
|
+
return 0.;
|
815
819
|
sparse_ix *search_res = std::lower_bound(row_st, row_end, (sparse_ix) col_num);
|
816
820
|
if (search_res == row_end || *search_res != (sparse_ix)col_num)
|
817
|
-
return 0
|
821
|
+
return 0.;
|
818
822
|
else
|
819
823
|
return prediction_data.Xr[search_res - prediction_data.Xr_ind];
|
820
824
|
}
|
@@ -22,7 +22,7 @@
|
|
22
22
|
* [9] Cortes, David. "Imputing missing values with unsupervised random trees." arXiv preprint arXiv:1911.06646 (2019).
|
23
23
|
*
|
24
24
|
* BSD 2-Clause License
|
25
|
-
* Copyright (c)
|
25
|
+
* Copyright (c) 2020, David Cortes
|
26
26
|
* All rights reserved.
|
27
27
|
* Redistribution and use in source and binary forms, with or without
|
28
28
|
* modification, are permitted provided that the following conditions are met:
|
@@ -0,0 +1,545 @@
|
|
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) 2020, 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
|
+
#include "isotree.hpp"
|
46
|
+
|
47
|
+
/* Translate isolation forest model into a single SQL select statement
|
48
|
+
*
|
49
|
+
* Parameters
|
50
|
+
* ==========
|
51
|
+
* - model_outputs
|
52
|
+
* Pointer to fitted single-variable model object from function 'fit_iforest'. Pass NULL
|
53
|
+
* if the predictions are to be made from an extended model. Can only pass one of
|
54
|
+
* 'model_outputs' and 'model_outputs_ext'.
|
55
|
+
* - model_outputs_ext
|
56
|
+
* Pointer to fitted extended model object from function 'fit_iforest'. Pass NULL
|
57
|
+
* if the predictions are to be made from a single-variable model. Can only pass one of
|
58
|
+
* 'model_outputs' and 'model_outputs_ext'.
|
59
|
+
* - table_from
|
60
|
+
* Table name from where the columns used in the model will be selected.
|
61
|
+
* - select_as
|
62
|
+
* Alias to give to the outlier score in the select statement.
|
63
|
+
* - numeric_colnames
|
64
|
+
* Names to use for the numerical columns.
|
65
|
+
* - categ_colnames
|
66
|
+
* Names to use for the categorical columns.
|
67
|
+
* - categ_levels
|
68
|
+
* Names to use for the levels/categories of each categorical column. These will be enclosed
|
69
|
+
* in single quotes.
|
70
|
+
* - index1
|
71
|
+
* Whether to make the node numbers start their numeration at 1 instead of 0 in the
|
72
|
+
* resulting statement. If passing 'output_tree_num=false', this will only affect the
|
73
|
+
* commented lines which act as delimiters. If passing 'output_tree_num=true', will also
|
74
|
+
* affect the results (which will also start at 1).
|
75
|
+
* - nthreads
|
76
|
+
* Number of parallel threads to use. Note that, the more threads, the more memory will be
|
77
|
+
* allocated, even if the thread does not end up being used. Ignored when not building with
|
78
|
+
* OpenMP support.
|
79
|
+
*
|
80
|
+
* Returns
|
81
|
+
* =======
|
82
|
+
* A string with the corresponding SQL statement that will calculate the outlier score
|
83
|
+
* from the model.
|
84
|
+
*/
|
85
|
+
std::string generate_sql_with_select_from(IsoForest *model_outputs, ExtIsoForest *model_outputs_ext,
|
86
|
+
std::string &table_from, std::string &select_as,
|
87
|
+
std::vector<std::string> &numeric_colnames, std::vector<std::string> &categ_colnames,
|
88
|
+
std::vector<std::vector<std::string>> &categ_levels,
|
89
|
+
bool index1, int nthreads)
|
90
|
+
{
|
91
|
+
std::vector<std::string> tree_conds = generate_sql(model_outputs, model_outputs_ext,
|
92
|
+
numeric_colnames, categ_colnames,
|
93
|
+
categ_levels,
|
94
|
+
false, index1, false, 0,
|
95
|
+
nthreads);
|
96
|
+
std::string out = std::accumulate(tree_conds.begin(), tree_conds.end(), std::string("SELECT\nPOWER(2.0, -(0.0"),
|
97
|
+
[&tree_conds, &index1](std::string &a, std::string &b)
|
98
|
+
{return a
|
99
|
+
+ std::string(" + \n---BEGIN TREE ")
|
100
|
+
+ std::to_string((size_t)std::distance(tree_conds.data(), &b) + (size_t)index1)
|
101
|
+
+ std::string("---\n")
|
102
|
+
+ b
|
103
|
+
+ std::string("\n---END OF TREE ")
|
104
|
+
+ std::to_string((size_t)std::distance(tree_conds.data(), &b) + (size_t)index1)
|
105
|
+
+ std::string("---\n");});
|
106
|
+
size_t ntrees = (model_outputs != NULL)? (model_outputs->trees.size()) : (model_outputs_ext->hplanes.size());
|
107
|
+
return
|
108
|
+
out
|
109
|
+
+ std::string(") / ")
|
110
|
+
+ std::to_string((long double)ntrees * ((model_outputs != NULL)?
|
111
|
+
(model_outputs->exp_avg_depth) : (model_outputs_ext->exp_avg_depth)))
|
112
|
+
+ std::string(") AS ")
|
113
|
+
+ select_as
|
114
|
+
+ std::string("\nFROM ")
|
115
|
+
+ table_from;
|
116
|
+
}
|
117
|
+
|
118
|
+
/* Translate model trees into SQL select statements
|
119
|
+
*
|
120
|
+
* Parameters
|
121
|
+
* ==========
|
122
|
+
* - model_outputs
|
123
|
+
* Pointer to fitted single-variable model object from function 'fit_iforest'. Pass NULL
|
124
|
+
* if the predictions are to be made from an extended model. Can only pass one of
|
125
|
+
* 'model_outputs' and 'model_outputs_ext'.
|
126
|
+
* - model_outputs_ext
|
127
|
+
* Pointer to fitted extended model object from function 'fit_iforest'. Pass NULL
|
128
|
+
* if the predictions are to be made from a single-variable model. Can only pass one of
|
129
|
+
* 'model_outputs' and 'model_outputs_ext'.
|
130
|
+
* - numeric_colnames
|
131
|
+
* Names to use for the numerical columns.
|
132
|
+
* - categ_colnames
|
133
|
+
* Names to use for the categorical columns.
|
134
|
+
* - categ_levels
|
135
|
+
* Names to use for the levels/categories of each categorical column. These will be enclosed
|
136
|
+
* in single quotes.
|
137
|
+
* - output_tree_num
|
138
|
+
* Whether to output the terminal node number instead of the separation depth at each node.
|
139
|
+
* - index1
|
140
|
+
* Whether to make the node numbers start their numeration at 1 instead of 0 in the
|
141
|
+
* resulting statement. If passing 'output_tree_num=false', this will only affect the
|
142
|
+
* commented lines which act as delimiters. If passing 'output_tree_num=true', will also
|
143
|
+
* affect the results (which will also start at 1).
|
144
|
+
* - single_tree
|
145
|
+
* Whether to generate the select statement for a single tree of the model instead of for
|
146
|
+
* all. The tree number to generate is to be passed under 'tree_num'.
|
147
|
+
* - tree_num
|
148
|
+
* Tree number for which to generate an SQL select statement, if passing 'single_tree=true'.
|
149
|
+
* - nthreads
|
150
|
+
* Number of parallel threads to use. Note that, the more threads, the more memory will be
|
151
|
+
* allocated, even if the thread does not end up being used. Ignored when not building with
|
152
|
+
* OpenMP support.
|
153
|
+
*
|
154
|
+
* Returns
|
155
|
+
* =======
|
156
|
+
* A vector containing at each element the SQL statement for the corresponding tree in the model.
|
157
|
+
* If passing 'single_tree=true', will contain only one element, corresponding to the tree given
|
158
|
+
* in 'tree_num'. The statements will be node-by-node, with commented-out separators using '---'
|
159
|
+
* as delimiters and including the node number as part of the comment.
|
160
|
+
*/
|
161
|
+
std::vector<std::string> generate_sql(IsoForest *model_outputs, ExtIsoForest *model_outputs_ext,
|
162
|
+
std::vector<std::string> &numeric_colnames, std::vector<std::string> &categ_colnames,
|
163
|
+
std::vector<std::vector<std::string>> &categ_levels,
|
164
|
+
bool output_tree_num, bool index1, bool single_tree, size_t tree_num,
|
165
|
+
int nthreads)
|
166
|
+
{
|
167
|
+
bool output_score = !output_tree_num;
|
168
|
+
size_t ntrees_use = single_tree?
|
169
|
+
1 : ((model_outputs != NULL)?
|
170
|
+
model_outputs->trees.size() : model_outputs_ext->hplanes.size());
|
171
|
+
std::string initial_str = std::string("\tWHEN\n");
|
172
|
+
|
173
|
+
size_t_for loop_st = 0;
|
174
|
+
size_t_for loop_end = ntrees_use;
|
175
|
+
if (single_tree)
|
176
|
+
{
|
177
|
+
loop_st = tree_num;
|
178
|
+
loop_end = loop_st + 1;
|
179
|
+
}
|
180
|
+
|
181
|
+
/* determine maximum number of nodes in a tree */
|
182
|
+
size_t max_nodes = 0;
|
183
|
+
for (size_t tree = loop_st; tree < loop_end; tree++)
|
184
|
+
max_nodes = std::max(max_nodes,
|
185
|
+
(model_outputs != NULL)?
|
186
|
+
(model_outputs->trees[tree].size()) : (model_outputs_ext->hplanes[tree].size()));
|
187
|
+
std::vector<std::string> conditions_left(max_nodes);
|
188
|
+
std::vector<std::string> conditions_right(max_nodes);
|
189
|
+
|
190
|
+
std::vector<std::vector<std::string>> all_node_rules(ntrees_use);
|
191
|
+
std::vector<std::string> out(ntrees_use);
|
192
|
+
|
193
|
+
size_t tree_use;
|
194
|
+
|
195
|
+
#pragma omp parallel for schedule(dynamic) num_threads(nthreads) \
|
196
|
+
shared(model_outputs, model_outputs_ext, numeric_colnames, categ_colnames, categ_levels, \
|
197
|
+
loop_st, loop_end, index1, single_tree, all_node_rules, out) \
|
198
|
+
firstprivate(conditions_left, conditions_right) private(tree_use)
|
199
|
+
for (size_t_for tree = loop_st; tree < loop_end; tree++)
|
200
|
+
{
|
201
|
+
if (model_outputs != NULL)
|
202
|
+
{
|
203
|
+
for (size_t node = 0; node < model_outputs->trees[tree].size(); node++)
|
204
|
+
extract_cond_isotree(*model_outputs, model_outputs->trees[tree][node],
|
205
|
+
conditions_left[node], conditions_right[node],
|
206
|
+
numeric_colnames, categ_colnames,
|
207
|
+
categ_levels);
|
208
|
+
}
|
209
|
+
|
210
|
+
else
|
211
|
+
{
|
212
|
+
for (size_t node = 0; node < model_outputs_ext->hplanes[tree].size(); node++)
|
213
|
+
extract_cond_ext_isotree(*model_outputs_ext, model_outputs_ext->hplanes[tree][node],
|
214
|
+
conditions_left[node], conditions_right[node],
|
215
|
+
numeric_colnames, categ_colnames,
|
216
|
+
categ_levels);
|
217
|
+
}
|
218
|
+
|
219
|
+
generate_tree_rules(
|
220
|
+
(model_outputs == NULL)? (NULL) : &(model_outputs->trees[tree]),
|
221
|
+
(model_outputs_ext == NULL)? (NULL) : &(model_outputs_ext->hplanes[tree]),
|
222
|
+
output_score,
|
223
|
+
0, index1, initial_str, all_node_rules[single_tree? 0 : tree],
|
224
|
+
conditions_left, conditions_right
|
225
|
+
);
|
226
|
+
|
227
|
+
/* Code below doesn't compile with MSVC (stuck with an OMP standard that's >20 years old) */
|
228
|
+
// if (single_tree)
|
229
|
+
// tree = 0;
|
230
|
+
tree_use = single_tree? (size_t)0 : tree;
|
231
|
+
|
232
|
+
if (all_node_rules[tree_use].size() <= 1)
|
233
|
+
{
|
234
|
+
for (std::string &rule : all_node_rules[tree_use])
|
235
|
+
rule = std::string("WHEN TRUE THEN ")
|
236
|
+
+ std::to_string((model_outputs != NULL)?
|
237
|
+
(model_outputs->exp_avg_depth) : (model_outputs_ext->exp_avg_depth))
|
238
|
+
+ std::string(" ");
|
239
|
+
}
|
240
|
+
|
241
|
+
out[tree_use] = std::accumulate(all_node_rules[tree_use].begin(), all_node_rules[tree_use].end(),
|
242
|
+
std::string("CASE\n"),
|
243
|
+
[&all_node_rules, &tree_use, &index1](std::string &a, std::string &b)
|
244
|
+
{return a
|
245
|
+
+ std::string("---begin terminal node ")
|
246
|
+
+ std::to_string((size_t)std::distance(&(all_node_rules[tree_use][0]), &b) + (size_t)index1)
|
247
|
+
+ std::string("---\n")
|
248
|
+
+ b;})
|
249
|
+
+ std::string("END\n");
|
250
|
+
all_node_rules[tree_use].clear();
|
251
|
+
}
|
252
|
+
|
253
|
+
return out;
|
254
|
+
}
|
255
|
+
|
256
|
+
|
257
|
+
void generate_tree_rules(std::vector<IsoTree> *trees, std::vector<IsoHPlane> *hplanes, bool output_score,
|
258
|
+
size_t curr_ix, bool index1, std::string &prev_cond, std::vector<std::string> &node_rules,
|
259
|
+
std::vector<std::string> &conditions_left, std::vector<std::string> &conditions_right)
|
260
|
+
{
|
261
|
+
if ((trees != NULL && (*trees)[curr_ix].score >= 0) ||
|
262
|
+
(hplanes != NULL && (*hplanes)[curr_ix].score >= 0))
|
263
|
+
{
|
264
|
+
node_rules.push_back(prev_cond
|
265
|
+
+ std::string("\tTHEN ")
|
266
|
+
+ (output_score?
|
267
|
+
(std::to_string((trees != NULL)?
|
268
|
+
((*trees)[curr_ix].score) : ((*hplanes)[curr_ix].score)))
|
269
|
+
:
|
270
|
+
(std::to_string(node_rules.size() + (size_t)index1)))
|
271
|
+
+ std::string("\n---end of terminal node ")
|
272
|
+
+ std::to_string(node_rules.size() + (size_t)index1)
|
273
|
+
+ std::string("---\n"));
|
274
|
+
return;
|
275
|
+
}
|
276
|
+
|
277
|
+
|
278
|
+
std::string cond_left = prev_cond
|
279
|
+
+ ((curr_ix > 0)? std::string("\t\tAND (") : std::string("\t\t ("))
|
280
|
+
+ conditions_left[curr_ix]
|
281
|
+
+ std::string(")\n");
|
282
|
+
generate_tree_rules(trees, hplanes, output_score,
|
283
|
+
(trees != NULL)?
|
284
|
+
((*trees)[curr_ix].tree_left) : ((*hplanes)[curr_ix].hplane_left),
|
285
|
+
index1, cond_left, node_rules,
|
286
|
+
conditions_left, conditions_right);
|
287
|
+
cond_left.clear();
|
288
|
+
std::string cond_right = prev_cond
|
289
|
+
+ ((curr_ix > 0)? std::string("\t\tAND (") : std::string("\t\t ("))
|
290
|
+
+ conditions_right[curr_ix]
|
291
|
+
+ std::string(")\n");
|
292
|
+
generate_tree_rules(trees, hplanes, output_score,
|
293
|
+
(trees != NULL)?
|
294
|
+
((*trees)[curr_ix].tree_right) : ((*hplanes)[curr_ix].hplane_right),
|
295
|
+
index1, cond_right, node_rules,
|
296
|
+
conditions_left, conditions_right);
|
297
|
+
}
|
298
|
+
|
299
|
+
|
300
|
+
void extract_cond_isotree(IsoForest &model, IsoTree &tree,
|
301
|
+
std::string &cond_left, std::string &cond_right,
|
302
|
+
std::vector<std::string> &numeric_colnames, std::vector<std::string> &categ_colnames,
|
303
|
+
std::vector<std::vector<std::string>> &categ_levels)
|
304
|
+
{
|
305
|
+
cond_left = std::string("");
|
306
|
+
cond_right = std::string("");
|
307
|
+
if (tree.score >= 0.)
|
308
|
+
return;
|
309
|
+
|
310
|
+
switch(tree.col_type)
|
311
|
+
{
|
312
|
+
case Numeric:
|
313
|
+
{
|
314
|
+
cond_left = ((model.missing_action != Impute)? (std::string("")) :
|
315
|
+
((tree.pct_tree_left >= .5)?
|
316
|
+
(numeric_colnames[tree.col_num]
|
317
|
+
+ std::string(" IS NULL OR "))
|
318
|
+
:
|
319
|
+
(numeric_colnames[tree.col_num]
|
320
|
+
+ std::string(" IS NOT NULL AND "))))
|
321
|
+
+ numeric_colnames[tree.col_num]
|
322
|
+
+ std::string(" <= ")
|
323
|
+
+ std::to_string(tree.num_split);
|
324
|
+
cond_right = ((model.missing_action != Impute)? (std::string("")) :
|
325
|
+
((tree.pct_tree_left >= .5)?
|
326
|
+
(numeric_colnames[tree.col_num]
|
327
|
+
+ std::string(" IS NOT NULL AND "))
|
328
|
+
:
|
329
|
+
(numeric_colnames[tree.col_num]
|
330
|
+
+ std::string(" IS NULL OR "))))
|
331
|
+
+ numeric_colnames[tree.col_num]
|
332
|
+
+ std::string(" > ")
|
333
|
+
+ std::to_string(tree.num_split);
|
334
|
+
break;
|
335
|
+
}
|
336
|
+
|
337
|
+
case Categorical:
|
338
|
+
{
|
339
|
+
switch(model.cat_split_type)
|
340
|
+
{
|
341
|
+
case SingleCateg:
|
342
|
+
{
|
343
|
+
cond_left = ((model.missing_action != Impute)? (std::string("")) :
|
344
|
+
((model.missing_action == Impute && tree.pct_tree_left >= .5)?
|
345
|
+
(categ_colnames[tree.col_num]
|
346
|
+
+ std::string(" IS NULL OR "))
|
347
|
+
:
|
348
|
+
(categ_colnames[tree.col_num]
|
349
|
+
+ std::string(" IS NOT NULL AND "))))
|
350
|
+
+ categ_colnames[tree.col_num]
|
351
|
+
+ std::string(" = '")
|
352
|
+
+ categ_levels[tree.col_num][tree.chosen_cat]
|
353
|
+
+ std::string("'");
|
354
|
+
cond_right = ((model.missing_action != Impute)? (std::string("")) :
|
355
|
+
((model.missing_action == Impute && tree.pct_tree_left >= .5)?
|
356
|
+
(categ_colnames[tree.col_num]
|
357
|
+
+ std::string(" IS NOT NULL AND "))
|
358
|
+
:
|
359
|
+
(categ_colnames[tree.col_num]
|
360
|
+
+ std::string(" IS NULL OR "))))
|
361
|
+
+ categ_colnames[tree.col_num]
|
362
|
+
+ std::string(" != '")
|
363
|
+
+ categ_levels[tree.col_num][tree.chosen_cat]
|
364
|
+
+ std::string("'");
|
365
|
+
break;
|
366
|
+
}
|
367
|
+
|
368
|
+
case SubSet:
|
369
|
+
{
|
370
|
+
cond_left = categ_colnames[tree.col_num] + std::string(" IN (");
|
371
|
+
cond_right = cond_left;
|
372
|
+
if (model.missing_action == Impute)
|
373
|
+
{
|
374
|
+
if (tree.pct_tree_left >= .5)
|
375
|
+
{
|
376
|
+
cond_left = categ_colnames[tree.col_num] + std::string(" IS NULL OR ") + cond_left;
|
377
|
+
cond_right = categ_colnames[tree.col_num] + std::string(" IS NOT NULL AND ") + cond_right;
|
378
|
+
}
|
379
|
+
|
380
|
+
else
|
381
|
+
{
|
382
|
+
cond_left = categ_colnames[tree.col_num] + std::string(" IS NOT NULL AND ") + cond_left;
|
383
|
+
cond_right = categ_colnames[tree.col_num] + std::string(" IS NULL OR ") + cond_right;
|
384
|
+
}
|
385
|
+
}
|
386
|
+
bool added_left = false;
|
387
|
+
bool added_right = false;
|
388
|
+
for (size_t categ = 0; categ < tree.cat_split.size(); categ++)
|
389
|
+
{
|
390
|
+
switch(tree.cat_split[categ])
|
391
|
+
{
|
392
|
+
case 1:
|
393
|
+
{
|
394
|
+
cond_left
|
395
|
+
+=
|
396
|
+
((added_left)? (std::string(", ")) : (std::string("")))
|
397
|
+
+ std::string("'")
|
398
|
+
+ categ_levels[tree.col_num][categ]
|
399
|
+
+ std::string("'");
|
400
|
+
added_left = true;
|
401
|
+
break;
|
402
|
+
}
|
403
|
+
|
404
|
+
case 0:
|
405
|
+
{
|
406
|
+
cond_right
|
407
|
+
+=
|
408
|
+
((added_right)? (std::string(", ")) : (std::string("")))
|
409
|
+
+ std::string("'")
|
410
|
+
+ categ_levels[tree.col_num][categ]
|
411
|
+
+ std::string("'");
|
412
|
+
added_right = true;
|
413
|
+
break;
|
414
|
+
}
|
415
|
+
|
416
|
+
case -1:
|
417
|
+
{
|
418
|
+
if (model.new_cat_action == Smallest || model.missing_action == Impute)
|
419
|
+
{
|
420
|
+
if ((model.new_cat_action == Smallest && tree.pct_tree_left < .5) ||
|
421
|
+
(model.missing_action == Impute && tree.pct_tree_left >= .5))
|
422
|
+
{
|
423
|
+
cond_left
|
424
|
+
+=
|
425
|
+
((added_left)? (std::string(", ")) : (std::string("")))
|
426
|
+
+ std::string("'")
|
427
|
+
+ categ_levels[tree.col_num][categ]
|
428
|
+
+ std::string("'");
|
429
|
+
added_left = true;
|
430
|
+
}
|
431
|
+
else
|
432
|
+
{
|
433
|
+
cond_right
|
434
|
+
+=
|
435
|
+
((added_right)? (std::string(", ")) : (std::string("")))
|
436
|
+
+ std::string("'")
|
437
|
+
+ categ_levels[tree.col_num][categ]
|
438
|
+
+ std::string("'");
|
439
|
+
added_right = true;
|
440
|
+
}
|
441
|
+
}
|
442
|
+
break;
|
443
|
+
}
|
444
|
+
}
|
445
|
+
}
|
446
|
+
if (added_left)
|
447
|
+
cond_left += std::string(")");
|
448
|
+
else
|
449
|
+
cond_left = std::string("");
|
450
|
+
if (added_right)
|
451
|
+
cond_right += std::string(")");
|
452
|
+
else
|
453
|
+
cond_right = std::string("");
|
454
|
+
|
455
|
+
break;
|
456
|
+
}
|
457
|
+
}
|
458
|
+
break;
|
459
|
+
}
|
460
|
+
}
|
461
|
+
}
|
462
|
+
|
463
|
+
void extract_cond_ext_isotree(ExtIsoForest &model, IsoHPlane &hplane,
|
464
|
+
std::string &cond_left, std::string &cond_right,
|
465
|
+
std::vector<std::string> &numeric_colnames, std::vector<std::string> &categ_colnames,
|
466
|
+
std::vector<std::vector<std::string>> &categ_levels)
|
467
|
+
{
|
468
|
+
cond_left = std::string("");
|
469
|
+
cond_right = std::string("");
|
470
|
+
if (hplane.score >= 0.)
|
471
|
+
return;
|
472
|
+
|
473
|
+
std::string hplane_conds = std::string("");
|
474
|
+
|
475
|
+
size_t n_visited_numeric = 0;
|
476
|
+
size_t n_visited_categ = 0;
|
477
|
+
for (size_t ix = 0; ix < hplane.col_num.size(); ix++)
|
478
|
+
{
|
479
|
+
hplane_conds
|
480
|
+
+=
|
481
|
+
((hplane_conds.length())? (std::string(" + ")) : (std::string("")))
|
482
|
+
+ ((model.missing_action == Impute)? (std::string("COALESCE(")) : (std::string("")));
|
483
|
+
switch(hplane.col_type[ix])
|
484
|
+
{
|
485
|
+
case Numeric:
|
486
|
+
{
|
487
|
+
hplane_conds
|
488
|
+
+=
|
489
|
+
std::to_string(hplane.coef[n_visited_numeric])
|
490
|
+
+ std::string(" * (")
|
491
|
+
+ numeric_colnames[hplane.col_num[ix]]
|
492
|
+
+ ((hplane.mean[n_visited_numeric] >= 0.)? (std::string(" - ")) : (std::string(" - (")))
|
493
|
+
+ std::to_string(hplane.mean[n_visited_numeric])
|
494
|
+
+ ((hplane.mean[n_visited_numeric] >= 0.)? (std::string(")")) : (std::string("))")));
|
495
|
+
n_visited_numeric++;
|
496
|
+
break;
|
497
|
+
}
|
498
|
+
|
499
|
+
case Categorical:
|
500
|
+
{
|
501
|
+
switch(model.cat_split_type)
|
502
|
+
{
|
503
|
+
case SingleCateg:
|
504
|
+
{
|
505
|
+
hplane_conds
|
506
|
+
+=
|
507
|
+
std::string("CASE WHEN ")
|
508
|
+
+ categ_colnames[hplane.col_num[ix]]
|
509
|
+
+ std::string(" = '")
|
510
|
+
+ categ_levels[hplane.col_num[ix]][hplane.chosen_cat[n_visited_categ]]
|
511
|
+
+ std::string("' THEN ")
|
512
|
+
+ std::to_string(hplane.fill_new[n_visited_categ])
|
513
|
+
+ std::string(" ELSE 0.0 END");
|
514
|
+
break;
|
515
|
+
}
|
516
|
+
|
517
|
+
case SubSet:
|
518
|
+
{
|
519
|
+
hplane_conds += std::string("CASE ") + categ_colnames[hplane.col_num[ix]];
|
520
|
+
for (size_t categ = 0; categ < hplane.cat_coef[hplane.col_num[ix]].size(); categ++)
|
521
|
+
{
|
522
|
+
hplane_conds
|
523
|
+
+=
|
524
|
+
std::string(" WHEN '")
|
525
|
+
+ categ_levels[hplane.col_num[ix]][categ]
|
526
|
+
+ std::string("' THEN ")
|
527
|
+
+ std::to_string( hplane.cat_coef[hplane.col_num[ix]][categ]);
|
528
|
+
}
|
529
|
+
if (model.new_cat_action == Smallest)
|
530
|
+
hplane_conds += std::string(" ELSE ") + std::to_string(hplane.fill_new[n_visited_categ]);
|
531
|
+
hplane_conds += std::string(" END");
|
532
|
+
break;
|
533
|
+
}
|
534
|
+
}
|
535
|
+
n_visited_categ++;
|
536
|
+
break;
|
537
|
+
}
|
538
|
+
}
|
539
|
+
hplane_conds += ((model.missing_action == Impute)?
|
540
|
+
(std::string(", ") + std::to_string(hplane.fill_val[ix]) + std::string(")")) : (std::string("")));
|
541
|
+
}
|
542
|
+
|
543
|
+
cond_left = hplane_conds + std::string(" <= ") + std::to_string(hplane.split_point);
|
544
|
+
cond_right = hplane_conds + std::string(" > ") + std::to_string(hplane.split_point);
|
545
|
+
}
|