public void Populate(DiscreteChoiceModel m)
{
Items.Clear();
if(m != null)
foreach (Smoother s in m.Smoothers)
{
Items.Add(s);
SetSelected(Items.IndexOf(s), true);
}
foreach (Smoother available in Smoother.Available)
if (Items.Cast<Smoother>().Count(present => present.GetType().Equals(available.GetType())) == 0)
Items.Add(available);
}
public static IEnumerable <DiscreteChoiceModel> GetAll(bool excludeThoseCopiedForPredictions)
{
IFeatureExtractor fex;
Configuration.TryGetFeatureExtractor(typeof(FeatureBasedDCM), out fex);
List <DiscreteChoiceModel> models = new List <DiscreteChoiceModel>();
BinaryFormatter bf = new BinaryFormatter();
NpgsqlCommand cmd = DB.Connection.NewCommand("SELECT " + Columns.Select + " FROM " + Table);
NpgsqlDataReader reader = cmd.ExecuteReader();
while (reader.Read())
{
DiscreteChoiceModel model = bf.Deserialize(new MemoryStream(reader[Table + "_" + Columns.Model] as byte[])) as DiscreteChoiceModel;
if (!excludeThoseCopiedForPredictions || (!model.HasMadePredictions && !model.IsMakingAPrediction))
{
models.Add(model);
}
}
reader.Close();
DB.Connection.Return(cmd.Connection);
return(models);
}
public static void Evaluate(Prediction prediction, int plotWidth, int plotHeight)
{
List <Incident> newIncidents = new List <Incident>();
DateTime newIncidentsStart = prediction.PredictionStartTime;
if (prediction.MostRecentlyEvaluatedIncidentTime >= newIncidentsStart)
{
newIncidentsStart = prediction.MostRecentlyEvaluatedIncidentTime + new TimeSpan(0, 0, 0, 0, 1);
}
foreach (Incident i in Incident.Get(newIncidentsStart, prediction.PredictionEndTime, prediction.PredictionArea, prediction.Model.IncidentTypes.ToArray()))
{
newIncidents.Add(i);
}
if (newIncidents.Count == 0)
{
return;
}
List <Incident> oldIncidents = new List <Incident>();
DateTime oldIncidentsEnd = newIncidentsStart - new TimeSpan(0, 0, 0, 0, 1);
if (oldIncidentsEnd >= prediction.PredictionStartTime)
{
foreach (Incident i in Incident.Get(prediction.PredictionStartTime, oldIncidentsEnd, prediction.PredictionArea, prediction.Model.IncidentTypes.ToArray()))
{
oldIncidents.Add(i);
}
}
IEnumerable <Incident> incidents = oldIncidents.Union(newIncidents);
DiscreteChoiceModel model = prediction.Model;
long sliceTicks = -1;
if (model is TimeSliceDCM)
{
sliceTicks = (model as TimeSliceDCM).TimeSliceTicks;
}
prediction.AssessmentPlots.Clear();
prediction.SliceThreatCorrelation.Clear();
prediction.OverallCrimeThreatCorrelation = -1;
int sliceNum = 1;
Dictionary <long, Dictionary <string, int> > sliceLocationTrueCount = SurveillancePlot.GetSliceLocationTrueCount(incidents, prediction);
Dictionary <long, Dictionary <string, List <double> > > sliceLocationThreats = SurveillancePlot.GetSliceLocationThreats(prediction);
foreach (long slice in sliceLocationTrueCount.Keys.OrderBy(slice => slice))
{
if (sliceLocationThreats.ContainsKey(slice)) // no prediction points might have been generated in the case of a feature-based classifier with no features
{
string slicePlotTitle = prediction.Name;
if (sliceTicks > 0)
{
DateTime sliceStart = new DateTime(slice * sliceTicks);
DateTime sliceEnd = sliceStart + new TimeSpan(sliceTicks - 1);
slicePlotTitle += Environment.NewLine +
sliceStart.ToShortDateString() + " " + sliceStart.ToShortTimeString() + " - " +
Environment.NewLine +
sliceEnd.ToShortDateString() + " " + sliceEnd.ToShortTimeString();
}
Dictionary <string, List <PointF> > seriesPoints = new Dictionary <string, List <PointF> >();
string predictionSeriesName = sliceLocationTrueCount.Count > 1 ? "Slice " + sliceNum++ : "Prediction";
seriesPoints.Add(predictionSeriesName, SurveillancePlot.GetSurveillancePlotPoints(sliceLocationTrueCount[slice], sliceLocationThreats[slice], true, true));
seriesPoints.Add(OptimalSeriesName, SurveillancePlot.GetOptimalSurveillancePlotPoints(sliceLocationTrueCount[slice], sliceLocationThreats[slice], true, true));
prediction.AssessmentPlots.Add(new SurveillancePlot(slicePlotTitle, slice, seriesPoints, plotHeight, plotWidth, Plot.Format.JPEG, 2));
prediction.SliceThreatCorrelation.Add(slice, GetThreatCorrelation(sliceLocationThreats[slice], sliceLocationTrueCount[slice]));
}
}
if (sliceLocationTrueCount.Count > 1)
{
Dictionary <string, int> overallLocationTrueCount = SurveillancePlot.GetOverallLocationTrueCount(incidents, prediction);
Dictionary <string, List <double> > overallLocationThreats = SurveillancePlot.GetOverallLocationThreats(prediction);
Dictionary <string, List <PointF> > seriesPoints = new Dictionary <string, List <PointF> >();
seriesPoints.Add("Overall", SurveillancePlot.GetSurveillancePlotPoints(overallLocationTrueCount, overallLocationThreats, true, true));
seriesPoints.Add(OptimalSeriesName, SurveillancePlot.GetOptimalSurveillancePlotPoints(overallLocationTrueCount, overallLocationThreats, true, true));
prediction.AssessmentPlots.Add(new SurveillancePlot(prediction.Name, -1, seriesPoints, plotHeight, plotWidth, Plot.Format.JPEG, 2));
prediction.OverallCrimeThreatCorrelation = GetThreatCorrelation(overallLocationThreats, overallLocationTrueCount);
}
prediction.MostRecentlyEvaluatedIncidentTime = incidents.Max(i => i.Time);
prediction.ReleaseAllLazyLoadedData();
}
internal void CommitValues(DiscreteChoiceModel model)
{
model.Name = ModelName;
model.TrainingArea = TrainingArea;
model.TrainingStart = TrainingStart;
model.TrainingEnd = TrainingEnd;
model.IncidentTypes = IncidentTypes;
model.Smoothers = Smoothers;
}
