Extended Isolation Forest
C++ Implementation of Extended Isolation Forest.
This is an extended version of this standard isolated forest. Compared with the standard isolated forest, the efficiency of the algorithm is reduced, but it alleviates the problems mentioned in the paper.
For an N dimensional dataset, selecting a random slope for the branch cut is the same as choosing a normal vector, n⃗ , uniformly over the unit N-Sphere. This can easily be accomplished by drawing a random number for each coordinate of n⃗ from the standard normal distribution N(0,1).
Installation
Just include EIF.h in your project and you are ready to go.
Example
#include <iostream>
#include <stdint.h>
#include <random>
#include <vector>
#include <string>
#include "EIF.h"
#include <chrono>
int main(int argc, char* argv[])
{
std::mt19937 rng(12345);
std::vector<std::array<uint32_t, 4>> data;
// generate some random 4D datapoints
for (uint32_t i = 0; i < 1000; i++)
{
std::array<uint32_t, 4> temp;
for (uint32_t j = 0; j < 4; j++)
{
temp[j] = EIF::UniformRandomNumber<float>::GenerateNext(rng, 0, 1000);
}
data.push_back(temp);
}
// add a few anomalies
for (uint32_t i = 0; i < 5; i++)
{
std::array<uint32_t, 4> temp;
for (uint32_t j = 0; j < 4; j++)
{
temp[j] = EIF::UniformRandomNumber<float>::GenerateNext(rng, 900, 1500);
}
data.push_back(temp);
}
auto t1 = std::chrono::system_clock::now();
EIF::EIForest<uint32_t, 4> forest;
if (!forest.Build(50, 12345, data, 100))
{
std::cerr << "Failed to build Isolation Forest.\n";
return 1;
}
std::vector<double> anomaly_scores;
if (!forest.GetAnomalyScores(data, anomaly_scores))
{
std::cerr << "Failed to calculate anomaly scores.\n";
return 2;
}
for (uint32_t i = 995; i < 1005; i++)
{
std::cout << "Anomaly_score[" << i << "] " << anomaly_scores[i] << "\n";
}
auto t2 = std::chrono::system_clock::now();
std::cout<< "time: "<<std::chrono::duration_cast<std::chrono::milliseconds>(t2 - t1).count()<<std::endl;
}Example output
Anomaly_score[995] 0.556233
Anomaly_score[996] 0.446035
Anomaly_score[997] 0.523133
Anomaly_score[998] 0.458597
Anomaly_score[999] 0.532598
Anomaly_score[1000] 0.588844
Anomaly_score[1001] 0.641518
Anomaly_score[1002] 0.639913
Anomaly_score[1003] 0.626346
Anomaly_score[1004] 0.639888
time: 84 milliseconds