public void detect_no_anomalies_if_factors_are_set_to_max_value()
{
var strategy = new OnlineNormalStrategy(Double.MaxValue, Double.MaxValue);
var anomalyResult = strategy.Detect(data, (0, int.MaxValue));
var expected = Enumerable.Empty <(int, Anomaly)>();
anomalyResult.SequenceEqual(expected).ShouldBeTrue();
}
private VerificationResult CreateAnomalyChecksAndRunEverything(
DataFrame data,
IMetricsRepository repository,
Check otherCheck,
IEnumerable <IAnalyzer <IMetric> > additionalRequiredAnalyzers)
{
// We only want to use historic data with the EU tag for the anomaly checks since the new
// data point is from the EU marketplace
var filterEU = new Dictionary <string, string> {
{ "marketplace", "EU" }
};
// We only want to use data points before the date time associated with the current
// data point and only ones that are from 2018
var afterDateTime = CreateDate(2018, 1, 1);
var beforeDateTime = CreateDate(2018, 8, 1);
// Config for the size anomaly check
var sizeAnomalyCheckConfig = new AnomalyCheckConfig(CheckLevel.Error, "Size only increases",
filterEU, afterDateTime, beforeDateTime);
var sizeAnomalyDetectionStrategy = new AbsoluteChangeStrategy(0);
// Config for the mean sales anomaly check
var meanSalesAnomalyCheckConfig = new AnomalyCheckConfig(
CheckLevel.Warning,
"Sales mean within 2 standard deviations",
filterEU,
afterDateTime,
beforeDateTime
);
var meanSalesAnomalyDetectionStrategy = new OnlineNormalStrategy(upperDeviationFactor: 2, lowerDeviationFactor: Option <double> .None,
ignoreAnomalies: false);
// ResultKey to be used when saving the results of this run
var currentRunResultKey =
new ResultKey(CreateDate(2018, 8, 1), new Dictionary <string, string> {
{ "marketplace", "EU" }
});
return(new VerificationSuite()
.OnData(data)
.AddCheck(otherCheck)
.AddRequiredAnalyzers(additionalRequiredAnalyzers)
.UseRepository(repository)
// Add the Size anomaly check
.AddAnomalyCheck(sizeAnomalyDetectionStrategy, Initializers.Size(), sizeAnomalyCheckConfig)
// Add the Mean sales anomaly check
.AddAnomalyCheck(meanSalesAnomalyDetectionStrategy, Initializers.Mean("sales"),
meanSalesAnomalyCheckConfig)
// Save new data point in the repository after we calculated everything
.SaveOrAppendResult(currentRunResultKey)
.Run());
}
public void detect_all_anomalies_if_no_interval_specified()
{
var strategy = new OnlineNormalStrategy(3.5, 3.5, 0.2);
var anomalyResult = strategy.Detect(data, (0, int.MaxValue));
var expected = Enumerable.Range(20, 11)
.Select(value => (value, new Anomaly(data[value], 1.0)));
anomalyResult.SequenceEqual(expected).ShouldBeTrue();
}
public void doesn_t_ignore_anomalies_in_calculation_if_not_wanted()
{
var strategy = new OnlineNormalStrategy(1.5, 1.5, 0.2, false);
var data = new[] { 1.0, 1.0, 1.0, 2.0, 1.0, 1.0, 1.0 };
var lastPoint = strategy.ComputeStatsAndAnomalies(data, (0, int.MaxValue)).Last();
var mean = data.Average();
var stdDev = StdDev(data);
lastPoint.Mean.ShouldBe(mean);
(Math.Abs(lastPoint.StdDev - stdDev) < stdDev * 0.001).ShouldBeTrue();
}
public void ignore_upper_factor_if_none_is_given()
{
var strategy = new OnlineNormalStrategy(1.5, Option <double> .None);
var anomalyResult = strategy.Detect(data, (0, int.MaxValue));
var expected = Enumerable
.Range(21, 9)
.Where(x => x % 2 != 0)
.Select(value => (value, new Anomaly(data[value], 1.0)));
anomalyResult.SequenceEqual(expected).ShouldBeTrue();
}
public OnlineNormalStrategyTest()
{
_strategy = new OnlineNormalStrategy(1.5, 1.5, 0.2);
}
