-
Notifications
You must be signed in to change notification settings - Fork 1
/
svr_rain_predictor.cpp
136 lines (113 loc) · 3.69 KB
/
svr_rain_predictor.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
#include "Matrix.h"
#include "Math.h"
#include "csv.h"
#include <iostream>
#include <cmath>
#include <string>
#include <vector>
typedef std::pair<double, Matrix> P;
double similarilty(Matrix l, Matrix x, double sigma){ //gaussian kernal
Matrix diff = l-x;
double p = (diff.transpose() * diff)(0)/(2*sigma*sigma);
return pow(Math::exp,-p);
}
Matrix makeFeatureMatrix(Matrix X, double sigma=1, Matrix L = Matrix(0)){
if(!L.rows()) L = X;
Matrix f(X.rows(), L.rows());
for(int i=0, r=X.rows(), c = L.rows(); i<r; i++){
for(int j=0; j<c; j++){
f(i,j) = similarilty(X.getRow(i).transpose(), L.getRow(j).transpose(), sigma);
}
}
return f;
}
P cost(Matrix& theta, Matrix& X, Matrix& y, double lambda){
int m = y.rows();
double J = 0;
Matrix h = (X*theta-y);
Matrix h1(h.rows());
for(int i=0; i<h.rows(); i++){
h1(i) = h(i)*h(i);
}
//std::cout<<"cost matrix: ";
// h1.show();
J = Math::sum(h1);
Matrix theta2 = theta * theta.transpose();
double theta_sq_sum = theta2(0)-1;
J += (lambda/2)*theta_sq_sum;
J /=m;
Matrix grad = X.transpose() * (h-y);
for(int i=1; i<grad.rows(); i++){
grad(i) = (grad(i) + lambda*theta(i))/m;
}
return std::make_pair(J, grad);
}
Matrix gradientDescent(Matrix& X, Matrix& y, Matrix theta, double lambda =1, int iters=500){
double p_cost = cost(theta, X, y, lambda ).first;
//std::cout << "INitial cost: ";cost(theta, X, y, lambda ).second.show();
//std::cout <<"initia cost: " << p_cost<< " gradientDescent called with: X:-";
//X.show();
//y.show();
double alpha = 0.5;
Matrix p_theta = theta;
while(iters--){
P t = cost(theta, X, y, lambda );
if(p_cost < t.first){
theta = p_theta;
alpha /=2.0;
}
p_cost = t.first;
p_theta = theta;
//if(iters%250==0) std::cout << "Cost: " << t.first << ' ';
theta = theta - t.second*alpha;
if(iters%500==0){
std::cout << ".";
std::cout.flush();
}
}
std::cout << '\n';
return theta;
}
Matrix predict_rain(Matrix data){
Matrix predicted_values(1,12);
for(int month=0; month<12; month++){
Matrix X(data.rows()-1, 12), y(data.rows()-1, 1);
for(int i=0; i<data.rows()-1; i++){
for(int j=0; j<12; j++){
if(j==month){
y(i) = data(i+1,j);
}
X(i,j) = data(i,j);
}
}
Matrix F = X;
Matrix mean_X = Math::mean(X);
Matrix range_X = Math::range(X);
Matrix mean_y = Math::mean(y);
Matrix range_y = Math::range(y);
Math::scale(y);
Math::scale(F);
Matrix scaled_X = F;
F = makeFeatureMatrix(F,0.15); //make feature matrix
F = F.append(Matrix::ones(F.rows()),0);
Matrix theta = gradientDescent(F, y, Matrix::zeros(F.cols()), 0.1, 5000);
//predict the value
Matrix test = X.getRow(X.rows()-1);
for(int i=0, r=test.rows(), c=test.cols(); i<r; i++){
for(int j=0; j<c; j++){
test(i,j) = (test(i,j) - mean_X(0,j) )/range_X(0,j);
}
}
test = makeFeatureMatrix(test, 0.15, scaled_X);
test = test.append(Matrix::ones(test.rows()),0);
Matrix ans = test*theta;
predicted_values(0,month) = (ans(0,0)*range_y(0,0))+mean_y(0,0);
}
return predicted_values;
}
int main(){
Matrix data = Matrix("training.dat");
//std::unordered_map<std::string, Matrix> mat = read_csv("rainfall_in_india_1901-2015.csv");
//mat["JHARKHAND"].save("training.dat");
predict_rain(data).save("predicted.dat");
}