public void Load(string sFileName, double dTrainSetSize)
{
Dictionary <RecommenderSystem.DatasetType, Tuple <Users, Items> > splittedData;
Tuple <Users, Items> data = dataLoaderEngine.Load(sFileName);
users = data.Item1;
items = data.Item2;
this.dsSize = dataLoaderEngine.GetDataSetSize();
splittedData = dataUtils.Split(dTrainSetSize, dsSize, data, DatasetType.Test, DatasetType.Train);
trainUsers = splittedData[DatasetType.Train].Item1;
trainItems = splittedData[DatasetType.Train].Item2;
testUsers = splittedData[DatasetType.Test].Item1;
testItems = splittedData[DatasetType.Test].Item2;
double trainSize = Math.Round(dsSize * dTrainSetSize);
splittedData = dataUtils.Split(dTrainSetSize, trainSize, new Tuple <Users, Items>(this.trainUsers, this.trainItems), DatasetType.Validation, DatasetType.Train);
trainUsers = splittedData[DatasetType.Train].Item1;
trainItems = splittedData[DatasetType.Train].Item2;
validationUsers = splittedData[DatasetType.Validation].Item1;
validationItems = splittedData[DatasetType.Validation].Item2;
//calculate the overall average rating
CalculateAverageRatingForTrainingSet();
similarityEngine = new SimilarityEngine(trainUsers, MAX_SIMILAR_USERS, logger); //TODO - check whether it should be train/test users
evaluationEngine = new EvaluationEngine(averageTrainRating);
predictionEngine.addModel(PredictionMethod.Cosine, new CollaborativeFilteringModel(trainUsers, trainItems, similarityEngine, new CosineMethod()));
predictionEngine.addModel(PredictionMethod.Pearson, new CollaborativeFilteringModel(trainUsers, trainItems, similarityEngine, new PearsonMethod()));
predictionEngine.addModel(PredictionMethod.Random, new CollaborativeFilteringModel(trainUsers, trainItems, similarityEngine, new RandomMethod()));
itemBasedEngine = new ItemBasedEngine(trainItems); //TODO - check that we need to use test items...
}
