forked from lballabio/QuantLib
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathlinearleastsquaresregression.cpp
247 lines (197 loc) · 8.4 KB
/
linearleastsquaresregression.cpp
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
/* -*- mode: c++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*- */
/*
Copyright (C) 2006 Klaus Spanderen
Copyright (C) 2010 Slava Mazur
This file is part of QuantLib, a free-software/open-source library
for financial quantitative analysts and developers - http://quantlib.org/
QuantLib is free software: you can redistribute it and/or modify it
under the terms of the QuantLib license. You should have received a
copy of the license along with this program; if not, please email
<quantlib-dev@lists.sf.net>. The license is also available online at
<http://quantlib.org/license.shtml>.
This program is distributed in the hope that it will be useful, but WITHOUT
ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
FOR A PARTICULAR PURPOSE. See the license for more details.
*/
#include "toplevelfixture.hpp"
#include "utilities.hpp"
#include <ql/math/randomnumbers/rngtraits.hpp>
#include <ql/math/linearleastsquaresregression.hpp>
#include <boost/circular_buffer.hpp>
#include <functional>
using namespace QuantLib;
using namespace boost::unit_test_framework;
BOOST_FIXTURE_TEST_SUITE(QuantLibTests, TopLevelFixture)
BOOST_AUTO_TEST_SUITE(LinearLeastSquaresRegressionTests)
BOOST_AUTO_TEST_CASE(testRegression) {
BOOST_TEST_MESSAGE("Testing linear least-squares regression...");
const Real tolerance = 0.05;
const Size nr=100000;
PseudoRandom::rng_type rng(PseudoRandom::urng_type(1234U));
std::vector<std::function<Real(Real)>> v = {
[](Real x) -> Real { return 1.0; },
[](Real x) -> Real { return x; },
[](Real x) -> Real { return x*x; },
[](Real x) -> Real { return std::sin(x); }
};
std::vector<std::function<Real(Real)>> w(v);
w.emplace_back([](Real x){ return x*x; });
for (Size k=0; k<3; ++k) {
Size i;
const Real a[] = {
rng.next().value,
rng.next().value,
rng.next().value,
rng.next().value
};
std::vector<Real> x(nr), y(nr);
for (i=0; i<nr; ++i) {
x[i] = rng.next().value;
// regression in y = a_1 + a_2*x + a_3*x^2 + a_4*sin(x) + eps
y[i] = a[0]*v[0](x[i]) + a[1]*v[1](x[i]) + a[2]*v[2](x[i])
+ a[3]*v[3](x[i]) + rng.next().value;
}
LinearRegression m(x, y, v);
for (i=0; i<v.size(); ++i) {
if (m.standardErrors()[i] > tolerance) {
BOOST_ERROR("Failed to reproduce linear regression coef."
<< "\n error: " << m.standardErrors()[i]
<< "\n tolerance: " << tolerance);
}
if (std::fabs(m.coefficients()[i]-a[i]) > 3*m.standardErrors()[i]) {
BOOST_ERROR("Failed to reproduce linear regression coef."
<< "\n calculated: " << m.coefficients()[i]
<< "\n error: " << m.standardErrors()[i]
<< "\n expected: " << a[i]);
}
}
m = LinearRegression(x, y, w);
const Real ma[] = {m.coefficients()[0], m.coefficients()[1],
m.coefficients()[2]+m.coefficients()[4],
m.coefficients()[3]};
const Real err[] = {m.standardErrors()[0], m.standardErrors()[1],
std::sqrt( m.standardErrors()[2]*m.standardErrors()[2]
+m.standardErrors()[4]*m.standardErrors()[4]),
m.standardErrors()[3]};
for (i=0; i<v.size(); ++i) {
if (std::fabs(ma[i] - a[i]) > 3*err[i]) {
BOOST_ERROR("Failed to reproduce linear regression coef."
<< "\n calculated: " << ma[i]
<< "\n error: " << err[i]
<< "\n expected: " << a[i]);
}
}
}
}
struct get_item {
Size i;
explicit get_item(Size i) : i(i) {}
Real operator()(const Array& a) const {
return a[i];
}
};
BOOST_AUTO_TEST_CASE(testMultiDimRegression) {
BOOST_TEST_MESSAGE(
"Testing multi-dimensional linear least-squares regression...");
const Size nr=100000;
const Size dims = 4;
const Real tolerance = 0.01;
PseudoRandom::rng_type rng(PseudoRandom::urng_type(1234U));
std::vector<std::function<Real(Array)> > v;
v.emplace_back([](const Array& x) { return 1.0; });
for (Size i=0; i < dims; ++i) {
v.emplace_back(get_item(i));
}
Array coeff(v.size());
for (Size i=0; i < v.size(); ++i) {
coeff[i] = rng.next().value;
}
std::vector<Real> y(nr, 0.0);
std::vector<Array> x(nr, Array(dims));
for (Size i=0; i < nr; ++i) {
for (Size j=0; j < dims; ++j) {
x[i][j] = rng.next().value;
}
for (Size j=0; j < v.size(); ++j) {
y[i] += coeff[j]*v[j](x[i]);
}
y[i] += rng.next().value;
}
LinearRegression m(x, y, v);
for (Size i=0; i < v.size(); ++i) {
if (m.standardErrors()[i] > tolerance) {
BOOST_ERROR("Failed to reproduce linear regression coef."
<< "\n error: " << m.standardErrors()[i]
<< "\n tolerance: " << tolerance);
}
if (std::fabs(m.coefficients()[i]-coeff[i]) > 3*tolerance) {
BOOST_ERROR("Failed to reproduce linear regression coef."
<< "\n calculated: " << m.coefficients()[i]
<< "\n error: " << m.standardErrors()[i]
<< "\n expected: " << coeff[i]);
}
}
// much simpler
LinearRegression m1(x, y, Real(1.0));
for (Size i=0; i < m1.dim(); ++i) {
if (m1.standardErrors()[i] > tolerance) {
BOOST_ERROR("Failed to reproduce linear regression coef."
<< "\n error: " << m1.standardErrors()[i]
<< "\n tolerance: " << tolerance);
}
if (std::fabs(m1.coefficients()[i]-coeff[i]) > 3*tolerance) {
BOOST_ERROR("Failed to reproduce linear regression coef."
<< "\n calculated: " << m1.coefficients()[i]
<< "\n error: " << m1.standardErrors()[i]
<< "\n expected: " << coeff[i]);
}
}
}
BOOST_AUTO_TEST_CASE(test1dLinearRegression) {
BOOST_TEST_MESSAGE("Testing 1D simple linear least-squares regression...");
/* Example taken from the QuantLib-User list, see posting
* Multiple linear regression/weighted regression, Boris Skorodumov */
std::vector<Real> x = {2.4, 1.8, 2.5, 3.0, 2.1, 1.2, 2.0, 2.7, 3.6};
std::vector<Real> y = {7.8, 5.5, 8.0, 9.0, 6.5, 4.0, 6.3, 8.4, 10.2};
std::vector<std::function<Real(Real)>> v = {
[](Real x) { return 1.0; },
[](Real x) { return x; }
};
LinearRegression m(x, y);
const Real tol = 0.0002;
const Real coeffExpected[] = { 0.9448, 2.6853 };
const Real errorsExpected[] = { 0.3654, 0.1487 };
for (Size i=0; i < 2; ++i) {
if (std::fabs(m.standardErrors()[i]-errorsExpected[i]) > tol) {
BOOST_ERROR("Failed to reproduce linear regression standard errors"
<< "\n calculated: " << m.standardErrors()[i]
<< "\n expected: " << errorsExpected[i]
<< "\n tolerance: " << tol);
}
if (std::fabs(m.coefficients()[i]-coeffExpected[i]) > tol) {
BOOST_ERROR("Failed to reproduce linear regression coef."
<< "\n calculated: " << m.coefficients()[i]
<< "\n expected: " << coeffExpected[i]
<< "\n tolerance: " << tol);
}
}
// an alternative container type
boost::circular_buffer<Real> cx(x.begin(), x.end()), cy(y.begin(), y.end());
LinearRegression m1(cx, cy);
for (Size i=0; i < 2; ++i) {
if (std::fabs(m1.standardErrors()[i]-errorsExpected[i]) > tol) {
BOOST_ERROR("Failed to reproduce linear regression standard errors"
<< "\n calculated: " << m1.standardErrors()[i]
<< "\n expected: " << errorsExpected[i]
<< "\n tolerance: " << tol);
}
if (std::fabs(m1.coefficients()[i]-coeffExpected[i]) > tol) {
BOOST_ERROR("Failed to reproduce linear regression coef."
<< "\n calculated: " << m1.coefficients()[i]
<< "\n expected: " << coeffExpected[i]
<< "\n tolerance: " << tol);
}
}
}
BOOST_AUTO_TEST_SUITE_END()
BOOST_AUTO_TEST_SUITE_END()