public void setupEscenario1()
{
Dictionary <Customer, DataPoint> mapCusToDP = new Dictionary <Customer, DataPoint>();
Dictionary <int, Item> mapDimToIt = new Dictionary <int, Item>();
mapCusToDP.Add(new Customer("1", "alv", "alv", "alv", "alv"), new DataPoint(1, new double[] { 2, 8, 4, 0, 6, 7, 0 }));
mapCusToDP.Add(new Customer("2", "alv", "alv", "alv", "alv"), new DataPoint(2, new double[] { 0, 2, 5, 6, 1, 8, 0 }));
mapCusToDP.Add(new Customer("3", "alv", "alv", "alv", "alv"), new DataPoint(3, new double[] { 2, 2, 3, 9, 0, 5, 3 }));
mapCusToDP.Add(new Customer("4", "alv", "alv", "alv", "alv"), new DataPoint(4, new double[] { 2, 8, 4, 0, 6, 7, 0 }));
mapCusToDP.Add(new Customer("5", "alv", "alv", "alv", "alv"), new DataPoint(5, new double[] { 2, 0, 4, 0, 6, 3, 0 }));
mapCusToDP.Add(new Customer("6", "alv", "alv", "alv", "alv"), new DataPoint(6, new double[] { 5, 0, 2, 0, 6, 0, 9 }));
mapCusToDP.Add(new Customer("7", "alv", "alv", "alv", "alv"), new DataPoint(7, new double[] { 2, 3, 4, 5, 0, 7, 0 }));
mapCusToDP.Add(new Customer("8", "alv", "alv", "alv", "alv"), new DataPoint(8, new double[] { 0, 9, 0, 0, 6, 7, 1 }));
mapDimToIt.Add(0, new Item("a", "lv"));
mapDimToIt.Add(1, new Item("b", "lv"));
mapDimToIt.Add(2, new Item("c", "lv"));
mapDimToIt.Add(3, new Item("d", "lv"));
mapDimToIt.Add(4, new Item("e", "lv"));
mapDimToIt.Add(5, new Item("f", "lv"));
mapDimToIt.Add(6, new Item("g", "lv"));
kMeans = new SimilarityAnalysisKMeans(4, 3, 7, 9, mapCusToDP, mapDimToIt);
}
private void setupEscenario2()
{
kMeans = new SimilarityAnalysisKMeans(4, 3, 7, 9, new Dictionary <Customer, DataPoint>(), new Dictionary <int, Item>());
}
public List <Recommendation> GetItemsCustomersMightBuyMoreButBuyFew(int numberOfGroups, int numberOfItemsInAnalysis, int minimumNumberOfItemPerCustomer, int numberOfItemsToRecommend)
{
int numberOfIterations = 100;
SimilarityAnalysisKMeans sakm = new SimilarityAnalysisKMeans(data, numberOfItemsInAnalysis, numberOfGroups, numberOfIterations, minimumNumberOfItemPerCustomer);
sakm.Kmeans();
List <List <Tuple <Customer, double[]> > > groups = sakm.GetCustomers();
List <Recommendation> recommendations = new List <Recommendation>();
int groupNum = 0;
foreach (List <Tuple <Customer, double[]> > group in groups)
{
double[] average = new double[sakm.dimensionOfDataPoints];
foreach (Tuple <Customer, double[]> items in group)
{
double[] vec = items.Item2;
for (int i = 0; i < vec.Length; i++)
{
average[i] += vec[i];
}
}
for (int i = 0; i < average.Length; i++)
{
average[i] = average[i] / group.Count;
}
foreach (Tuple <Customer, double[]> customer in group)
{
List <Tuple <Item, double> > diffs = new List <Tuple <Item, double> >();
for (int i = 0; i < customer.Item2.GetLength(0); i++)
{
diffs.Add(new Tuple <Item, double>(sakm.mapFromDimensionToItem[i], average[i] - customer.Item2[i]));
}
diffs = diffs.OrderByDescending(t => t.Item2).Take(numberOfItemsToRecommend).ToList();
Recommendation re = new Recommendation(diffs);
re.customerRepresentation = customer.Item2;
re.customer = customer.Item1;
re.groupColor = groupNum;
recommendations.Add(re);
}
groupNum++;
}
recommendations.Sort();
double[][] matrix = new double[recommendations.Count][];
for (int i = 0; i < recommendations.Count; i++)
{
double[] dat = recommendations[i].customerRepresentation;
matrix[i] = dat;
}
PrincipalComponentAnalysis pca = new PrincipalComponentAnalysis(matrix);
pca.NumberOfOutputs = 2;
// Compute the Principal Component Analysis
pca.Compute();
// Creates a projection
double[][] components = pca.Transform(matrix);
for (int i = 0; i < components.GetLength(0); i++)
{
recommendations[i].customer2dRepresentation = components[i];
}
recommendations = recommendations.OrderBy(r => r.customer.id).ToList();
return(recommendations);
}
