1 | #ifndef _theplu_yat_utility_nni_ |
---|
2 | #define _theplu_yat_utility_nni_ |
---|
3 | |
---|
4 | // $Id: NNI.h 1271 2008-04-09 16:11:07Z peter $ |
---|
5 | |
---|
6 | /* |
---|
7 | Copyright (C) 2004 Jari Häkkinen |
---|
8 | Copyright (C) 2005, 2006 Jari Häkkinen, Peter Johansson |
---|
9 | Copyright (C) 2007 Peter Johansson |
---|
10 | Copyright (C) 2008 Jari Häkkinen, Peter Johansson |
---|
11 | |
---|
12 | This file is part of the yat library, http://trac.thep.lu.se/yat |
---|
13 | |
---|
14 | The yat library is free software; you can redistribute it and/or |
---|
15 | modify it under the terms of the GNU General Public License as |
---|
16 | published by the Free Software Foundation; either version 2 of the |
---|
17 | License, or (at your option) any later version. |
---|
18 | |
---|
19 | The yat library is distributed in the hope that it will be useful, |
---|
20 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
---|
21 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
---|
22 | General Public License for more details. |
---|
23 | |
---|
24 | You should have received a copy of the GNU General Public License |
---|
25 | along with this program; if not, write to the Free Software |
---|
26 | Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA |
---|
27 | 02111-1307, USA. |
---|
28 | */ |
---|
29 | |
---|
30 | #include "Matrix.h" |
---|
31 | |
---|
32 | #include <iostream> |
---|
33 | #include <utility> |
---|
34 | #include <vector> |
---|
35 | |
---|
36 | namespace theplu { |
---|
37 | namespace yat { |
---|
38 | namespace utility { |
---|
39 | |
---|
40 | /// |
---|
41 | /// @brief Interface class for nearest |
---|
42 | /// neighbour imputation (NNI) algorithms. |
---|
43 | /// |
---|
44 | /// NNI algorithms implemented here is discussed in documents |
---|
45 | /// created in the WeNNI project. This document will be released for |
---|
46 | /// public access, and the necessary information for retrieving that |
---|
47 | /// document will be provided here. |
---|
48 | /// |
---|
49 | /// Short introduction to NNI is that one may want to improve |
---|
50 | /// (correct) uncertain data. Here, the data to be imputed is stored in a |
---|
51 | /// matrix where rows similar to each other are used to adjust |
---|
52 | /// uncertain data. The data matrix is accompanied by a weight |
---|
53 | /// (uncertainty) matrix defining what data is to be considered as |
---|
54 | /// 'certain' and what data is uncertain. The weight matrix can be |
---|
55 | /// binary with 1's indicating that the data does not need |
---|
56 | /// corrections, whereas a 0 means that the data should be replaced |
---|
57 | /// by an imputed value. Naturally, the weight matrix can also be |
---|
58 | /// continuous where values between 0 and 1 defines how certain a |
---|
59 | /// data element is. |
---|
60 | /// |
---|
61 | /// The imputation depends on how similarity of rows of data is |
---|
62 | /// defined and on the number of closest neighbours (here; rows) to |
---|
63 | /// use in the imputation can be set. |
---|
64 | /// |
---|
65 | /// Implementation issues |
---|
66 | /// |
---|
67 | /// The current implementation treats rows where all data are tagged |
---|
68 | /// are completely uncertain, i.e. all weights are zero, by |
---|
69 | /// ignoring these lines in nearest neighbourhood |
---|
70 | /// calculations. Importantly, this type of data are not changed |
---|
71 | /// (imputed) either since there is no close neighbourhood defined |
---|
72 | /// for this data. |
---|
73 | /// |
---|
74 | /// Rows that is completely identical in an imputation algorithm |
---|
75 | /// sense will give problems since the distance between will usually |
---|
76 | /// become zero. This is solved by setting zero distance to a small |
---|
77 | /// number. Identical rows in this context are basically a |
---|
78 | /// comparison between elements with non-zero uncertainty weights |
---|
79 | /// only, and all these elements are equal. Zero weight elements are |
---|
80 | /// not used in the comparison since these are considered as |
---|
81 | /// non/sense values. |
---|
82 | /// |
---|
83 | class NNI |
---|
84 | { |
---|
85 | public: |
---|
86 | |
---|
87 | /// |
---|
88 | /// Base constructor for the nearest neighbour imputation |
---|
89 | /// algorithms. |
---|
90 | /// |
---|
91 | NNI(const utility::Matrix& matrix,const utility::Matrix& weight, |
---|
92 | const unsigned int neighbours); |
---|
93 | |
---|
94 | virtual ~NNI(void) {}; |
---|
95 | |
---|
96 | /// |
---|
97 | /// Function doing the imputation. |
---|
98 | /// |
---|
99 | /// @return number of rows not imputed |
---|
100 | /// |
---|
101 | virtual unsigned int estimate(void)=0; |
---|
102 | |
---|
103 | /// |
---|
104 | /// @return A const reference to the modified data. |
---|
105 | /// |
---|
106 | const utility::Matrix& imputed_data(void) const; |
---|
107 | |
---|
108 | /// |
---|
109 | /// @return indices of rows in data matrix not imputed |
---|
110 | /// |
---|
111 | const std::vector<size_t>& not_imputed(void) const; |
---|
112 | |
---|
113 | protected: |
---|
114 | /** |
---|
115 | \f$ d_{ij}^2=\frac {\sum_{k=1}^C w_{ik} w_{jk} (x_{ik}-x_{jk})^2 |
---|
116 | }{\sum_{k=l}^C w_{ik} w_{jk} } \f$ where C is the number of columns |
---|
117 | */ |
---|
118 | std::vector<std::pair<size_t,double> > |
---|
119 | calculate_distances(const size_t) const; |
---|
120 | /// Contributing nearest neighbours are added up to the user set |
---|
121 | /// number, and neighbours are disqualified if their element |
---|
122 | /// (column) weight is zero |
---|
123 | std::vector<size_t> |
---|
124 | nearest_neighbours(const size_t, |
---|
125 | const std::vector<std::pair<size_t,double> >&) const; |
---|
126 | /// |
---|
127 | /// original data matrix |
---|
128 | /// |
---|
129 | const utility::Matrix& data_; |
---|
130 | |
---|
131 | /// |
---|
132 | /// data after imputation |
---|
133 | /// |
---|
134 | utility::Matrix imputed_data_; |
---|
135 | |
---|
136 | /// |
---|
137 | /// number of neighbor to use |
---|
138 | /// |
---|
139 | unsigned int neighbours_; |
---|
140 | |
---|
141 | /// |
---|
142 | /// which rows are not imputed due to lack of data |
---|
143 | /// |
---|
144 | std::vector<size_t> not_imputed_; |
---|
145 | |
---|
146 | /// |
---|
147 | /// weight matrix |
---|
148 | /// |
---|
149 | const utility::Matrix& weight_; |
---|
150 | }; |
---|
151 | |
---|
152 | }}} // of namespace utility, yat, and theplu |
---|
153 | |
---|
154 | #endif |
---|