1 | // $Id: ROC.h 148 2004-09-09 13:22:41Z peter $ |
---|
2 | |
---|
3 | #ifndef _theplu_cpptools_roc_ |
---|
4 | #define _theplu_cpptools_roc_ |
---|
5 | |
---|
6 | // C++ tools include |
---|
7 | ///////////////////// |
---|
8 | #include "Score.h" |
---|
9 | #include "vector.h" |
---|
10 | |
---|
11 | // Standard C++ includes |
---|
12 | //////////////////////// |
---|
13 | #include <utility> |
---|
14 | #include <vector> |
---|
15 | |
---|
16 | namespace theplu { |
---|
17 | namespace cpptools { |
---|
18 | /// |
---|
19 | /// Class for ROC (Reciever Operating Characteristic). |
---|
20 | /// |
---|
21 | |
---|
22 | class ROC : public Score |
---|
23 | { |
---|
24 | |
---|
25 | public: |
---|
26 | /// |
---|
27 | /// Default constructor |
---|
28 | /// |
---|
29 | ROC(); |
---|
30 | |
---|
31 | /// |
---|
32 | /// Destructor |
---|
33 | /// |
---|
34 | virtual ~ROC(void) {}; |
---|
35 | |
---|
36 | /// Function taking \a value, \a target (+1 or -1) and vector |
---|
37 | /// defining what samples to use. The score is equivalent to |
---|
38 | /// Mann-Whitney statistics @return the area under the ROC |
---|
39 | /// curve. If the area is less than 0.5, is 1-area returned. |
---|
40 | /// |
---|
41 | double score(const gslapi::vector& value, const gslapi::vector& target, |
---|
42 | const std::vector<size_t>& = std::vector<size_t>()); |
---|
43 | |
---|
44 | /// Function taking values, target, weight and a vector defining |
---|
45 | /// what samples to use. The area is defines as \f$ \frac{\sum |
---|
46 | /// w^+w^-}{\sum w^+w^-}\f$, where the sum in the numerator goes |
---|
47 | /// over all pairs where value+ is larger than value-. The |
---|
48 | /// denominator goes over all pairs. @return wheighted version of |
---|
49 | /// area under the ROC curve. If the area is less than 0.5, is |
---|
50 | /// 1-area returned. |
---|
51 | /// |
---|
52 | double score(const gslapi::vector&, const gslapi::vector&, |
---|
53 | const gslapi::vector&, |
---|
54 | const std::vector<size_t>& = std::vector<size_t>()); |
---|
55 | |
---|
56 | /// |
---|
57 | ///Calculates the p-value, i.e. the probability of observing an area |
---|
58 | ///equally or larger if the null hypothesis is true. If P is near zero, |
---|
59 | ///this casts doubt on this hypothesis. The null hypothesis is that the |
---|
60 | ///values from the 2 classes are generated from 2 identical |
---|
61 | ///distributions. The alternative is that the median of the first |
---|
62 | ///distribution is shifted from the median of the second distribution by a |
---|
63 | ///non-zero amount. If the smallest group size is larger than minimum_size |
---|
64 | ///(default = 10), then P is calculated using a normal approximation. |
---|
65 | /// @return the one-sided p-value |
---|
66 | /// |
---|
67 | double p_value() ; |
---|
68 | |
---|
69 | /// |
---|
70 | /// @return the targets in train_set sorted with respect to the |
---|
71 | /// corresponding data |
---|
72 | /// |
---|
73 | gslapi::vector ROC::target(void) const; |
---|
74 | |
---|
75 | /// |
---|
76 | /// Changes minimum_size , i.e. the threshold when a normal |
---|
77 | /// approximation is used for the p-value calculation. |
---|
78 | /// |
---|
79 | inline void minimum_size(const u_int minimum_size) |
---|
80 | {minimum_size_ = minimum_size; } |
---|
81 | |
---|
82 | private: |
---|
83 | double area_; |
---|
84 | gslapi::vector data_; |
---|
85 | u_int minimum_size_; |
---|
86 | u_int nof_pos_; |
---|
87 | gslapi::vector target_; |
---|
88 | std::vector<size_t> train_set_; |
---|
89 | std::vector<std::pair<double, double> > value_; |
---|
90 | /// pair of target and data. should always be sorted with respect to |
---|
91 | /// data. |
---|
92 | gslapi::vector weight_; |
---|
93 | |
---|
94 | /// |
---|
95 | /// |
---|
96 | /// Implemented as in MatLab 13.1 |
---|
97 | /// @return the p-value |
---|
98 | /// |
---|
99 | double ROC::get_p_approx(const double) const; |
---|
100 | |
---|
101 | /// |
---|
102 | /// @return the p-value |
---|
103 | /// |
---|
104 | double ROC::get_p_exact(const double, const double, |
---|
105 | const double); |
---|
106 | |
---|
107 | /// |
---|
108 | /// sorting value_, should always be done when changing train_set_ |
---|
109 | /// |
---|
110 | void ROC::sort(); |
---|
111 | |
---|
112 | }; |
---|
113 | |
---|
114 | /// |
---|
115 | /// The output operator for the ROC class. The output is an Nx2 |
---|
116 | /// matrix, where the first column is the sensitivity and second |
---|
117 | /// is the specificity. |
---|
118 | /// |
---|
119 | std::ostream& operator<< (std::ostream& s, const ROC&); |
---|
120 | |
---|
121 | |
---|
122 | }} // of namespace cpptools and namespace theplu |
---|
123 | |
---|
124 | #endif |
---|
125 | |
---|