// div3Sigma sets given metric score and average via 3-sigma.
// states that nearly all values (99.7%) lie within the 3 standard deviations
// of the mean in a normal distribution.
func (d *Detector) div3Sigma(m *models.Metric, vals []float64) {
if len(vals) == 0 {
m.Score = 0
m.Average = m.Value
return
}
// Values average and standard deviation.
avg := mathutil.Average(vals)
std := mathutil.StdDev(vals, avg)
// Set metric average
m.Average = avg
// Set metric score
if len(vals) <= int(d.cfg.Detector.LeastCount) { // Number of values not enough
m.Score = 0
return
}
last := m.Value
if std == 0 { // Eadger
switch {
case last == avg:
m.Score = 0
case last > avg:
m.Score = 1
case last < avg:
m.Score = -1
}
return
}
m.Score = (last - avg) / (3 * std) // 3-sigma
}
// Aggregate QueryCost.
func aggregationQueryCost() {
h.info.lock.Lock()
defer h.info.lock.Unlock()
h.queryCostsLock.Lock()
defer h.queryCostsLock.Unlock()
h.info.QueryCost = mathutil.Average(h.queryCosts)
h.queryCosts = h.queryCosts[:0]
}
// Aggregate FilterCost.
func aggregationFilterCost() {
h.info.lock.Lock()
defer h.info.lock.Unlock()
h.filterCostsLock.Lock()
defer h.filterCostsLock.Unlock()
h.info.FilterCost = mathutil.Average(h.filterCosts)
h.filterCosts = h.filterCosts[:0]
}
// Aggregate DetectionCost.
func aggregateDetectionCost() {
h.info.lock.Lock()
defer h.info.lock.Unlock()
h.detectionCostsLock.Lock()
defer h.detectionCostsLock.Unlock()
h.info.DetectionCost = mathutil.Average(h.detectionCosts)
h.detectionCosts = h.detectionCosts[:0]
}