public List <KeyValuePair <int, double> > predict(UserPreferences targetUser, List <ScoreClass> neighbours)
{
Dictionary <int, double> predictedRatings = new Dictionary <int, double>();
List <int> articlesToCheck = new List <int>();
foreach (var neighbour in neighbours)
{
foreach (var rating in neighbour.User.Prefs)
{
if (!targetUser.Prefs.ContainsKey(rating.Key) && !articlesToCheck.Contains(rating.Key))
{
articlesToCheck.Add(rating.Key);
}
}
}
while (articlesToCheck.Any())
{
var article = articlesToCheck.Last();
double sumWeightedRatings = 0;
double sumCoefficients = 0;
var neighboursWithArticle = 0;
foreach (var neighbour in neighbours)
{
if (neighbour.User.Prefs.ContainsKey(article))
{
neighboursWithArticle++;
sumWeightedRatings += (neighbour.User.Prefs[article] * neighbour.Score);
sumCoefficients += neighbour.Score;
}
}
var predictedRating = sumWeightedRatings / sumCoefficients;
articlesToCheck.Remove(article);
if (neighboursWithArticle > 2)
{
predictedRatings.Add(article, predictedRating);
}
}
var sortedPredictions = predictedRatings.ToList();
sortedPredictions.Sort((firstPair, nextPair) =>
{
return(-firstPair.Value.CompareTo(nextPair.Value));
});
return(sortedPredictions);
}
public double Measure(UserPreferences user1, UserPreferences user2)
{
double x = 0;
foreach (var pair in user1.Prefs)
{
if (user2.Prefs.ContainsKey(pair.Key))
{
double pi = pair.Value;
double qi = user2.Prefs[pair.Key];
x += Math.Pow((pi - qi), 2);
}
}
double distance = Math.Round(Math.Sqrt(x), 3);
double score = 1 / (1 + distance);
return(score);
}
public double Measure(UserPreferences user1, UserPreferences user2)
{
double xy = 0;
double rootX = 0;
double rootY = 0;
foreach (var pair in user1.Prefs)
{
var a = pair.Value;
rootX += Math.Pow(Math.Abs(a), 2);
if (user2.Prefs.ContainsKey(pair.Key))
{
var b = user2.Prefs[pair.Key];
xy += (a * b);
rootY += Math.Pow(Math.Abs(b), 2);
}
}
var score = (xy / (Math.Sqrt(rootX) * Math.Sqrt(rootY)));
return(score);
}
public double Measure(UserPreferences user1, UserPreferences user2)
{
double sumXiYi = 0;
double sumXi = 0;
double sumYi = 0;
double sumXiPow = 0;
double sumYiPow = 0;
double n = 0;
foreach (var pair in user1.Prefs)
{
if (user2.Prefs.ContainsKey(pair.Key))
{
n += 1;
var xi = pair.Value;
var yi = user2.Prefs[pair.Key];
//upper part of formula
//left part
sumXiYi += (xi * yi);
//right part
sumXi += xi;
sumYi += yi;
//bottom part also uses sumXi and sumYi
//bottom left
sumXiPow += Math.Pow(xi, 2);
sumYiPow += Math.Pow(yi, 2);
}
}
var topPart = sumXiYi - ((sumXi * sumYi) / n);
var bottomLeftPart = Math.Sqrt(sumXiPow - (Math.Pow(sumXi, 2) / n));
var bottomRightPart = Math.Sqrt(sumYiPow - (Math.Pow(sumYi, 2) / n));
var score = Math.Round(topPart / (bottomLeftPart * bottomRightPart), 3);
return(score);
}
private static void ApplicationCode()
{
var dataSet = new Dictionary <int, UserPreferences>();
var nearestNeighboursFinder = new NearestNeighboursFinder();
var predictor = new Predictor();
var minimumThreshold = 0.35;
var targetUserId = 3;
var neighbourAmount = 3;
var recommendAmount = 3;
StreamReader reader = null;
////////////////////////////////////////////////
// load dataset from file with a filereader
// movielens.data = 100k dataset
// UserItem.data = default dataset
////////////////////////////////////////////////
var chooseDataset = SelectDataset();
if (chooseDataset.Equals("movielens"))
{
reader = new StreamReader(File.OpenRead("../../movielens.data"));
targetUserId = 186;
neighbourAmount = 25;
recommendAmount = 8;
}
else
{
reader = new StreamReader(File.OpenRead("../../UserItem.data"));
targetUserId = 7;
neighbourAmount = 3;
recommendAmount = 3;
}
while (!reader.EndOfStream)
{
var line = reader.ReadLine();
var values = line.Split(',');
var userId = Convert.ToInt32(values[0]);
var article = Convert.ToInt32(values[1]);
var rating = Convert.ToDouble(values[2], CultureInfo.InvariantCulture);
if (dataSet.ContainsKey(userId))
{
dataSet[userId].AddRating(article, rating);
}
else
{
var up = new UserPreferences(userId, article, rating);
dataSet.Add(userId, up);
}
}
//Comment / Uncomment to toggle writing all users and their ratings to the console
//foreach (var user in dataSet)
//{
// user.Value.PrintAllRatings();
//}
////////////////////////////////////////////////
// Write scores of comparing users to console //
////////////////////////////////////////////////
var algorithm = SelectAlgorithm();
var neighbours = nearestNeighboursFinder.Find(dataSet, targetUserId, minimumThreshold, algorithm, neighbourAmount);
Console.WriteLine("\n15 Nearest neighbours: \n");
foreach (var i in neighbours)
{
Console.WriteLine(i.User.OwnerId + " With score " + i.Score.ToString("0.0000"));
}
Console.WriteLine("");
//Make predictions based on target user with the set of neighbours from previous step.
List <KeyValuePair <int, double> > predictions = predictor.predict(dataSet[targetUserId], neighbours);
var suggestions = predictions.Take(recommendAmount);
foreach (var suggestion in suggestions)
{
Console.WriteLine("Recommended article " + suggestion.Key + " With score " + suggestion.Value.ToString("0.0000"));
}
// wait for R or Esc before restarting/exiting
Console.WriteLine("\npress Escape to quit the application or type R to restart");
restartOrNot();
}
