public virtual bool isAnomaly(Timeseries ts, correlatedFeatures cf, int timeStep)
{
List <float> Column1 = ts.GetColumn(cf.feature1);
List <float> Column2 = ts.GetColumn(cf.feature2);
Point currPoint = new Point(Column1[timeStep], Column2[timeStep]);
float currDev = Anomaly_Detection_Util.dev(currPoint, cf.lin_reg);
return(currDev > cf.threshold * 1.1);
}
public bool isAnomalous(correlatedFeatures cf, Point point)
{
float correlation = adu.dev(point, cf.lin_reg);
if (correlation > cf.threshold)
{
return(true);
}
return(false);
}
new public void cirCorr(correlatedFeatures cf, Point[] points, int size)
{
Circle minCircle = new Circle(new Point(0, 0), 0);
Circle circle = minCircle.findMinCircle(points, size);
cf.centerX = circle.center.x;
cf.centerY = circle.center.y;
cf.rad = circle.radius;
cf.threshold = (float)(cf.rad * 1.1);
cf.isCircle = true;
this.cf.Add(cf);
}
public override bool isAnomaly(Timeseries ts, correlatedFeatures cf, int timeStep)
{
if (cf.corrlation > circThreshold)
{
var Column1 = ts.GetColumn(cf.feature1);
var Column2 = ts.GetColumn(cf.feature2);
Point currPoint = new Point(Column1[timeStep], Column2[timeStep]);
//Point point = cf.center
return(Circle_Util.distanceBetweenPoints(currPoint, cf.center) > cf.threshold * 1.1);
}
return(false);
}
public virtual void addCorrelation(Timeseries ts, string feat1, string feat2, float pearson)
{
if (pearson > linThreshold)
{
correlatedFeatures correlation = new correlatedFeatures();
correlation.feature1 = feat1;
correlation.feature2 = feat2;
correlation.corrlation = pearson;
var Column1 = ts.GetColumn(feat1);
var Column2 = ts.GetColumn(feat2);
correlation.lin_reg = Anomaly_Detection_Util.LinReg(Column1, Column2);
correlation.threshold = getMaxDev(Timeseries.CombineColumns(Column1, Column2), correlation.lin_reg);
cf.Add(correlation);
}
}
new public bool isAnomalous(correlatedFeatures cf, Point point)
{
if (!cf.isCircle)
{
return(isAnomalous(cf, point));
}
else
{
float cor = (float)Math.Sqrt(Math.Pow(point.x - cf.centerX, 2) + Math.Pow(point.y - cf.centerY, 2));
if (cor > cf.threshold)
{
return(true);
}
}
return(false);
}
public override void addCorrelation(Timeseries ts, string feat1, string feat2, float pearson)
{
if (pearson > circThreshold)
{
correlatedFeatures correlation = new correlatedFeatures();
correlation.feature1 = feat1;
correlation.feature2 = feat2;
correlation.corrlation = pearson;
var Column1 = ts.GetColumn(feat1);
var Column2 = ts.GetColumn(feat2);
var points = Timeseries.CombineColumns(Column1, Column2);
Circle welzlCirc = Circle_Util.welzl(points, new List <Point>());
correlation.center = welzlCirc.center;
correlation.threshold = welzlCirc.radius;
cf.Add(correlation);
}
}
public void learnNormal(Timeseries ts)
{
anomaly_detection_util adu = new anomaly_detection_util();
Dictionary <string, List <float> > tsMap = ts.table;
List <string> tsFeaturesVector = ts.features;
int featuresVectorSize = tsFeaturesVector.Count();
int valueVectorSize = tsMap[tsFeaturesVector[0]].Count();
for (int i = 0; i < featuresVectorSize; ++i)
{
string fiName = tsFeaturesVector[i];
float[] fiData = tsMap[fiName].ToArray();
float maxPearson = 0;
correlatedFeatures currentCF = new correlatedFeatures();
currentCF.feature1 = fiName;
for (int j = i + 1; j < featuresVectorSize; ++j)
{
string fjName = tsFeaturesVector[j];
float[] fjData = tsMap[fjName].ToArray();
float currentPearson = Math.Abs(adu.pearson(fiData, fjData, valueVectorSize));
if (currentPearson > maxPearson)
{
maxPearson = currentPearson;
currentCF.feature2 = fjName;
}
}
if (maxPearson > this.threshold)
{
currentCF.corrlation = maxPearson;
Point[] points = new Point[valueVectorSize];
for (int j = 0; j < valueVectorSize; ++j)
{
points[j] = new Point(tsMap[currentCF.feature1][j], tsMap[currentCF.feature2][j]);
}
currentCF.lin_reg = adu.linear_reg(points, valueVectorSize);
currentCF.threshold = 0;
for (int j = 0; j < valueVectorSize; ++j)
{
currentCF.threshold = Math.Max(currentCF.threshold, Math.Abs(adu.dev(points[j], currentCF.lin_reg)));
}
currentCF.threshold *= (float)1.1;
cf.Add(currentCF);
}
}
}
public float cirCorr(correlatedFeatures cf, Point[] points, int size)
{
return(0);
}
