protected virtual StreamPipeline ComposeOffPipeline(string id, DataExchange exchange)
{
var offlinePipeline = new StreamPipeline();
Dictionary<string, int> time = new Dictionary<string, int>();
var trainUntil = testStartDateProvider.GetTimestampOfTestStart(id);
offlinePipeline.Register(new offPredictionFittingFilter(parameters, trainUntil, time, exchange));
return offlinePipeline;
}
protected virtual StreamPipeline ComposeOnPipeline(string id, DataExchange exchange, string name)
{
var onlinePipeline = new StreamPipeline();
parameters = new FilterParameters();
var trainUntil = testStartDateProvider.GetTimestampOfTestStart(id);
onlinePipeline.Register(new onNeuralPredictionFilter(parameters, trainUntil, exchange));
onlinePipeline.Register(new onErrorCalculationFilter(parameters));
onlinePipeline.Register(new ResultOutputFilter(repository) { MeasurementId = id, ForecastModelId = name});
return onlinePipeline;
}
public virtual PipelinePack ComposePipeline(string id, string name)
{
List<List<float>> DbRealValue = new List<List<float>>();
List<DateTime> DbDateTime = new List<DateTime>();
var exchange = new DataExchange();
var onlinePipeline = ComposeOnPipeline(id, exchange, name);
var offlinePipeline = ComposeOffPipeline(id, exchange);
var databasePipeline = ComposeDbPipeline();
return new PipelinePack { OnlinePipeline = onlinePipeline, OfflinePipeline = offlinePipeline, DatabasePipeline = databasePipeline };
}
public onNeuralPredictionFilter(FilterParameters parameters, DateTime trainTill, DataExchange exchange = null)
{
this.exchange = exchange;
this.waitUntil = trainTill;
this.timeSeries = new TimeSeries(0);
settings = new ForecastSettings();
for (int i = 0; i < 5; i++)
{
settings.energyLags.Add(i + 1);
}
model = new MultipleNeuralNetworksModel();
this.parameters = parameters;
parameters.Values["model"] = model;
}
public offPredictionFittingFilter(FilterParameters parameters, DateTime trainTill, Dictionary<string, int> time, DataExchange exchange = null)
{
this.exchange = exchange;
this.time = time;
this.waitUntil = trainTill;
this.parameters = parameters;
settings = new ForecastSettings();
for (int i = 0; i < 5; i++)
{
settings.energyLags.Add(i + 1);
}
for (int i = 0; i < PACKAGE_SIZE; i++)
{
timeSeriesEnsemble.Add(new TimeSeries(0));
}
}
