public void NNearestNeighbour_Evaluation_Using_AverageAbsoluteDifferenceMetric()
{
RandomUtils.useTestSeed();
AverageAbsoluteDifferenceRecommenderEvaluator evaluator = new AverageAbsoluteDifferenceRecommenderEvaluator();
IRecommenderBuilder pearsonRecommenderBuilder = new TestRecommenderBuilderPearson();
IRecommenderBuilder euclideanRecommenderBuilder = new TestRecommenderBuilderEuclidean();
IRecommenderBuilder logLikelyRecommenderBuilder = new TestRecommenderBuilderLogLikelihood();
IRecommenderBuilder spearmanRecommenderBuilder = new TestRecommenderBuilderSpearman();
IRecommenderBuilder tanimotoRecommenderBuilder = new TestRecommenderBuilderTanimoto();
// Average Absolute Difference For the recommendations based on metric applied
// Common metric used for assessing the accuracy of the recommendations.
// The lower the result the better the algorithm.
// Training Data: 70%
// Test Data: 30%
using (var scope = Container.BeginLifetimeScope())
{
_recommenderService = scope.Resolve <RecommenderService>();
GenericDataModel model = _recommenderService.GetUserBasedDataModel();
double score = evaluator.Evaluate(pearsonRecommenderBuilder
, null, model, 0.7, 1.0);
double scoreEuclid = evaluator.Evaluate(euclideanRecommenderBuilder
, null, model, 0.7, 1.0);
double scoreLoglikely = evaluator.Evaluate(logLikelyRecommenderBuilder
, null, model, 0.7, 1.0);
double scoreSpearman = evaluator.Evaluate(spearmanRecommenderBuilder
, null, model, 0.7, 1.0);
double scoreTanimoto = evaluator.Evaluate(tanimotoRecommenderBuilder
, null, model, 0.7, 1.0);
}
}
public void TestEvaluate()
{
DataModel model = GetDataModel();
RecommenderBuilder builder = new SlopeOneRecommenderBuilder();
RecommenderEvaluator evaluator =
new AverageAbsoluteDifferenceRecommenderEvaluator();
double eval = evaluator.Evaluate(builder, model, 0.75, 1.0);
Assert.AreEqual(0.26387685767414826, eval, EPSILON);
}
