private static void startWRMF(ITimedRatings all_data)
{
removeUserThreshold(ref all_data);
Console.WriteLine("Start iteration Test WRMF");
//for (int i = 0; i < 5; i++) {
ITimedRatings validation_data = new TimedRatings(); // 10%
ITimedRatings test_data = new TimedRatings(); // 20%
ITimedRatings training_data = new TimedRatings(); // 70%
readAndSplitData(all_data, ref test_data, ref training_data, ref validation_data);
IPosOnlyFeedback training_data_pos = new PosOnlyFeedback <SparseBooleanMatrix> (); // 80%
for (int index = 0; index < training_data.Users.Count; index++)
{
training_data_pos.Add(training_data.Users [index], training_data.Items [index]);
}
MyMediaLite.ItemRecommendation.WRMF recommender = new MyMediaLite.ItemRecommendation.WRMF();
recommender.Feedback = training_data_pos;
DateTime start_time = DateTime.Now;
recommender.Train();
Console.Write("Total Training time needed:");
Console.WriteLine(((TimeSpan)(DateTime.Now - start_time)).TotalMilliseconds);
Console.WriteLine("Final results in this iteration:");
var results = MyMediaLite.Eval.ItemsWeatherItemRecommender.EvaluateTime(recommender, validation_data, training_data, "VALIDATION ", false);
results = MyMediaLite.Eval.ItemsWeatherItemRecommender.EvaluateTime(recommender, test_data, training_data, "TEST ", false);
//}
}
public void TestTimeSplit()
{
var ratings = new TimedRatings();
ratings.Add(0, 0, 5.0f, new DateTime(2011, 10, 31));
ratings.Add(0, 1, 4.5f, new DateTime(2011, 11, 1));
ratings.Add(1, 0, 1.0f, new DateTime(2011, 10, 31));
ratings.Add(1, 1, 2.5f, new DateTime(2011, 11, 2));
var split1 = new RatingsChronologicalSplit(ratings, new DateTime(2011, 11, 2));
Assert.AreEqual(3, split1.Train[0].Count);
Assert.AreEqual(1, split1.Test[0].Count);
Assert.AreEqual(2, split1.Train[0].ByUser[0].Count);
Assert.AreEqual(1, split1.Train[0].ByUser[1].Count);
Assert.AreEqual(0, split1.Test[0].ByUser[0].Count);
Assert.AreEqual(1, split1.Test[0].ByUser[1].Count);
Assert.AreEqual(new DateTime(2011, 10, 31), split1.Train[0].EarliestTime);
Assert.AreEqual(new DateTime(2011, 11, 1), split1.Train[0].LatestTime);
Assert.AreEqual(new DateTime(2011, 11, 2), split1.Test[0].EarliestTime);
Assert.AreEqual(new DateTime(2011, 11, 2), split1.Test[0].LatestTime);
var split2 = new RatingsChronologicalSplit(ratings, new DateTime(2011, 11, 1));
Assert.AreEqual(2, split2.Train[0].Count);
Assert.AreEqual(2, split2.Test[0].Count);
Assert.AreEqual(1, split2.Train[0].ByUser[0].Count);
Assert.AreEqual(1, split2.Train[0].ByUser[1].Count);
Assert.AreEqual(1, split2.Test[0].ByUser[0].Count);
Assert.AreEqual(1, split2.Test[0].ByUser[1].Count);
Assert.AreEqual(new DateTime(2011, 10, 31), split2.Train[0].EarliestTime);
Assert.AreEqual(new DateTime(2011, 10, 31), split2.Train[0].LatestTime);
Assert.AreEqual(new DateTime(2011, 11, 1), split2.Test[0].EarliestTime);
Assert.AreEqual(new DateTime(2011, 11, 2), split2.Test[0].LatestTime);
}
public void TestNumberSplit()
{
var ratings = new TimedRatings();
ratings.Add(0, 0, 5.0f, new DateTime(2011, 10, 31));
ratings.Add(0, 1, 4.5f, new DateTime(2011, 11, 1));
ratings.Add(0, 2, 5.0f, new DateTime(2011, 11, 3));
ratings.Add(0, 3, 4.5f, new DateTime(2011, 11, 4));
ratings.Add(1, 0, 1.0f, new DateTime(2011, 10, 31));
ratings.Add(1, 1, 2.5f, new DateTime(2011, 11, 2));
ratings.Add(1, 2, 1.0f, new DateTime(2011, 12, 1));
ratings.Add(1, 3, 2.5f, new DateTime(2011, 12, 4));
var split1 = new RatingsPerUserChronologicalSplit(ratings, 1);
Assert.AreEqual(6, split1.Train[0].Count);
Assert.AreEqual(2, split1.Test[0].Count);
Assert.AreEqual(3, split1.Train[0].ByUser[0].Count);
Assert.AreEqual(3, split1.Train[0].ByUser[1].Count);
Assert.AreEqual(1, split1.Test[0].ByUser[0].Count);
Assert.AreEqual(1, split1.Test[0].ByUser[1].Count);
var split2 = new RatingsPerUserChronologicalSplit(ratings, 2);
Assert.AreEqual(4, split2.Train[0].Count);
Assert.AreEqual(4, split2.Test[0].Count);
Assert.AreEqual(2, split2.Train[0].ByUser[0].Count);
Assert.AreEqual(2, split2.Train[0].ByUser[1].Count);
Assert.AreEqual(2, split2.Test[0].ByUser[0].Count);
Assert.AreEqual(2, split2.Test[0].ByUser[1].Count);
var split3 = new RatingsPerUserChronologicalSplit(ratings, 3);
Assert.AreEqual(2, split3.Train[0].Count);
Assert.AreEqual(6, split3.Test[0].Count);
Assert.AreEqual(1, split3.Train[0].ByUser[0].Count);
Assert.AreEqual(1, split3.Train[0].ByUser[1].Count);
Assert.AreEqual(3, split3.Test[0].ByUser[0].Count);
Assert.AreEqual(3, split3.Test[0].ByUser[1].Count);
var split4 = new RatingsPerUserChronologicalSplit(ratings, 4);
Assert.AreEqual(0, split4.Train[0].Count);
Assert.AreEqual(8, split4.Test[0].Count);
Assert.AreEqual(0, split4.Train[0].ByUser[0].Count);
Assert.AreEqual(0, split4.Train[0].ByUser[1].Count);
Assert.AreEqual(4, split4.Test[0].ByUser[0].Count);
Assert.AreEqual(4, split4.Test[0].ByUser[1].Count);
var split5 = new RatingsPerUserChronologicalSplit(ratings, 5);
Assert.AreEqual(0, split5.Train[0].Count);
Assert.AreEqual(8, split5.Test[0].Count);
Assert.AreEqual(0, split5.Train[0].ByUser[0].Count);
Assert.AreEqual(0, split5.Train[0].ByUser[1].Count);
Assert.AreEqual(4, split5.Test[0].ByUser[0].Count);
Assert.AreEqual(4, split5.Test[0].ByUser[1].Count);
}
private static void startIterationTestNew(ITimedRatings all_data, bool weather_aware, double beta, double mu, int city, int iterations, string feature = "")
{
removeUserThreshold(ref all_data);
Console.WriteLine("Start iteration test");
ITimedRatings validation_data = new TimedRatings(); // 10%
ITimedRatings test_data = new TimedRatings(); // 30%
ITimedRatings training_data = new TimedRatings(); // 70%
readAndSplitData(all_data, ref test_data, ref training_data, ref validation_data);
int K = 100;
int rangeSize = 10;
int user_count = all_data.AllUsers.Count;
int location_count = all_data.AllItems.Count();
double[,] U1 = new double[user_count, K];
double[,] U2 = new double[user_count, K];
double[,] L1 = new double[location_count, K];
double[,] L2 = new double[location_count, K];
double[,] L3 = new double[location_count, K];
double[,] F = new double[rangeSize, K];
Dictionary <int, int> idMapperCategories = new Dictionary <int, int> ();
Dictionary <int, int> idMapperLocations = new Dictionary <int, int> ();
Dictionary <int, int> idMapperUser = new Dictionary <int, int> ();
initMatrixNormal(all_data.AllUsers, ref U1, ref idMapperUser, K);
initMatrixNormal(all_data.AllUsers, ref U2, ref idMapperUser, K);
initMatrixNormal(all_data.AllItems, ref L1, ref idMapperLocations, K);
initMatrixNormal(all_data.AllItems, ref L2, ref idMapperLocations, K);
initMatrixNormal(all_data.AllItems, ref L3, ref idMapperLocations, K);
initMatrixNormal(rangeSize, ref F, K);
WeatherContextAwareItemRecommender recommender = new WeatherContextAwareItemRecommender(U1, U2, L1, L2, L3, F, idMapperLocations, idMapperUser, city, feature);
recommender.connection_string = connection;
recommender.Ratings = training_data;
recommender.Validation = validation_data;
recommender.Test = test_data;
recommender.weather_aware = weather_aware;
recommender.rangeSize = rangeSize;
recommender.max_iter = iterations;
recommender.evaluation_at = 20;
recommender.beta = beta;
DateTime start_time = DateTime.Now;
recommender.Train();
Console.Write("Total Training time needed:");
Console.WriteLine(((TimeSpan)(DateTime.Now - start_time)).TotalMilliseconds);
Console.WriteLine("Final results in this iteration:");
//}
}
public static IRatings CreateRandomTimedRatings(int num_users, int num_items, int num_ratings)
{
var random = MyMediaLite.Random.GetInstance();
var ratings = new TimedRatings();
for (int i = 0; i < num_ratings; i++)
{
int user_id = random.Next(num_users);
int item_id = random.Next(num_items);
int rating_value = 1 + random.Next(5);
ratings.Add(user_id, item_id, rating_value, DateTime.Now);
}
return(ratings);
}
private static void startIterationTest(ITimedRatings all_data, bool weather_aware, double beta, double mu, int city, int iterations)
{
removeUserThreshold(ref all_data);
Console.WriteLine("Start iteration test");
ITimedRatings validation_data = new TimedRatings(); // 10%
ITimedRatings test_data = new TimedRatings(); // 30%
ITimedRatings training_data = new TimedRatings(); // 70%
readAndSplitData(all_data, ref test_data, ref training_data, ref validation_data);
int K = 100;
int rangeSize = 10;
int user_count = all_data.AllUsers.Count;
int location_count = all_data.AllItems.Count();
double[,] U1 = new double[user_count, K];
double[,] U2 = new double[user_count, K];
double[,] U3 = new double[user_count, K];
double[,] L1 = new double[location_count, K];
Dictionary <int, int> idMapperLocations = new Dictionary <int, int> ();
Dictionary <int, int> idMapperUser = new Dictionary <int, int> ();
initMatrixNormal(all_data.AllUsers, ref U1, ref idMapperUser, K);
initMatrixNormal(all_data.AllUsers, ref U2, ref idMapperUser, K);
initMatrixNormal(all_data.AllUsers, ref U3, ref idMapperUser, K);
initMatrixNormal(all_data.AllItems, ref L1, ref idMapperLocations, K);
WeatherItemRecommender recommender = new WeatherItemRecommender(U1, U2, U3, L1, idMapperLocations, idMapperUser);
recommender.connection_string = connection;
recommender.Ratings = training_data;
recommender.Validation = validation_data;
recommender.Test = test_data;
recommender.weather_aware = weather_aware;
recommender.max_iter = iterations;
recommender.evaluation_at = 20;
recommender.beta = beta;
DateTime start_time = DateTime.Now;
recommender.Train();
Console.Write("Total Training time needed:");
Console.WriteLine(((TimeSpan)(DateTime.Now - start_time)).TotalMilliseconds);
Console.WriteLine("Final results in this iteration:");
var results = MyMediaLite.Eval.ItemsWeatherItemRecommender.EvaluateTime(recommender, validation_data, training_data, "VALIDATION ", false);
results = MyMediaLite.Eval.ItemsWeatherItemRecommender.EvaluateTime(recommender, test_data, training_data, "TEST ", false);
}
/// <summary>Read in rating data from a TextReader</summary>
/// <param name="reader">the <see cref="TextReader"/> to read from</param>
/// <param name="user_mapping">mapping object for user IDs</param>
/// <param name="item_mapping">mapping object for item IDs</param>
/// <param name="test_rating_format">whether there is a rating column in each line or not</param>
/// <returns>the rating data</returns>
static public ITimedRatings Read(
TextReader reader,
IMapping user_mapping = null, IMapping item_mapping = null,
TestRatingFileFormat test_rating_format = TestRatingFileFormat.WITH_RATINGS)
{
if (user_mapping == null)
{
user_mapping = new IdentityMapping();
}
if (item_mapping == null)
{
item_mapping = new IdentityMapping();
}
var ratings = new TimedRatings();
string[] separators = { "::" };
string line;
int seconds_pos = test_rating_format == TestRatingFileFormat.WITH_RATINGS ? 3 : 2;
while ((line = reader.ReadLine()) != null)
{
string[] tokens = line.Split(separators, StringSplitOptions.None);
if (test_rating_format == TestRatingFileFormat.WITH_RATINGS && tokens.Length < 4)
{
throw new FormatException("Expected at least 4 columns: " + line);
}
if (test_rating_format == TestRatingFileFormat.WITHOUT_RATINGS && tokens.Length < 3)
{
throw new FormatException("Expected at least 3 columns: " + line);
}
int user_id = user_mapping.ToInternalID(tokens[0]);
int item_id = item_mapping.ToInternalID(tokens[1]);
float rating = test_rating_format == TestRatingFileFormat.WITH_RATINGS ? float.Parse(tokens[2], CultureInfo.InvariantCulture) : 0;
long seconds = uint.Parse(tokens[seconds_pos]);
var time = new DateTime(seconds * 10000000L).AddYears(1969);
var offset = TimeZone.CurrentTimeZone.GetUtcOffset(time);
time -= offset;
ratings.Add(user_id, item_id, rating, time);
}
return(ratings);
}
public override void Train(Split split)
{
// Convert trainset to MyMediaLite trianset format
if (DataType == IO.DataType.Ratings)
{
var mmlFeedback = new Ratings();
foreach (var feedback in split.Train)
{
var rating = (Rating)feedback;
mmlFeedback.Add(UsersMap.ToInternalID(rating.User.Id), ItemsMap.ToInternalID(rating.Item.Id), rating.Value);
}
((IRatingPredictor)MmlRecommenderInstance).Ratings = mmlFeedback;
}
else if (DataType == IO.DataType.TimeAwareRating)
{
var mmlFeedback = new TimedRatings();
var firstRatingMl10M = new DateTime(1998, 11, 1);
foreach (var feedback in split.Train)
{
var rating = (Rating)feedback;
var time = firstRatingMl10M.AddDays(double.Parse(feedback.Attributes["timestamp"].Value));
mmlFeedback.Add(UsersMap.ToInternalID(rating.User.Id), ItemsMap.ToInternalID(rating.Item.Id),
rating.Value, time);
}
((ITimeAwareRatingPredictor)MmlRecommenderInstance).Ratings = mmlFeedback;
}
else
{
var mmlFeedback = new PosOnlyFeedback <SparseBooleanMatrix>();
foreach (var feedback in split.Train)
{
mmlFeedback.Add(UsersMap.ToInternalID(feedback.User.Id), ItemsMap.ToInternalID(feedback.Item.Id));
}
((ItemRecommender)MmlRecommenderInstance).Feedback = mmlFeedback;
if (MmlRecommenderInstance is IModelAwareRecommender)
{
((IModelAwareRecommender)MmlRecommenderInstance).Model = this;
}
}
Logger.Current.Trace("Training with MyMediaLite recommender...");
PureTrainTime = (int)Wrap.MeasureTime(delegate() { MmlRecommenderInstance.Train(); }).TotalMilliseconds;
}
private static void startUserKNN(string data)
{
MyMediaLite.Data.Mapping user_mapping = new MyMediaLite.Data.Mapping();
MyMediaLite.Data.Mapping item_mapping = new MyMediaLite.Data.Mapping();
ITimedRatings all_data = readDataMapped(data, ref user_mapping, ref item_mapping);
removeUserThreshold(ref all_data);
Console.WriteLine("Start iteration Test UserKNN");
//for (int i = 0; i < 5; i++) {
ITimedRatings validation_data = new TimedRatings(); // 10%
ITimedRatings test_data = new TimedRatings(); // 20%
ITimedRatings training_data = new TimedRatings(); // 70%
readAndSplitData(all_data, ref test_data, ref training_data, ref validation_data);
IPosOnlyFeedback training_data_pos = new PosOnlyFeedback <SparseBooleanMatrix> (); // 80%
for (int index = 0; index < training_data.Users.Count; index++)
{
training_data_pos.Add(training_data.Users [index], training_data.Items [index]);
}
MyMediaLite.ItemRecommendation.UserKNN recommender = new MyMediaLite.ItemRecommendation.UserKNN();
recommender.K = 80;
recommender.Q = 1;
recommender.Weighted = false;
recommender.Alpha = 0.5f;
recommender.Correlation = MyMediaLite.Correlation.BinaryCorrelationType.Jaccard;
recommender.Feedback = training_data_pos;
DateTime start_time = DateTime.Now;
recommender.Train();
Console.Write("Total Training time needed:");
Console.WriteLine(((TimeSpan)(DateTime.Now - start_time)).TotalMilliseconds);
Console.WriteLine("Final results in this iteration:");
var results = MyMediaLite.Eval.ItemsWeatherItemRecommender.EvaluateTime(recommender, validation_data, training_data, "VALIDATION ", false);
results = MyMediaLite.Eval.ItemsWeatherItemRecommender.EvaluateTime(recommender, test_data, training_data, "TEST ", false);
//}
}
// public static double startMuTuning(ITimedRatings all_data){
//
// ITimedRatings validation_data = new TimedRatings (); // 10%
// ITimedRatings test_data = new TimedRatings (); // 20%
// ITimedRatings training_data = new TimedRatings (); // 70%
// ITimedRatings sub_all_data = new TimedRatings();
//
// getSubset (all_data, ref sub_all_data);
// removeUserThreshold (ref sub_all_data);
// readAndSplitData (sub_all_data, ref validation_data, ref test_data, ref training_data);
// Dictionary<int,int> venueCategoryMapper = getCategories(sub_all_data);
//
// int K = 100;
// int user_count = sub_all_data.AllUsers.Count;
// int location_count = sub_all_data.AllItems.Count ();
// Console.WriteLine (location_count);
// int category_count = venueCategoryMapper.Values.ToList ().Distinct ().Count();
//
// double[,] U1 = new double[user_count,K];
// double[,] U2 = new double[user_count,K];
// double[,] U3 = new double[user_count,K];
// double[,] U4 = new double[user_count,K];
// double[,] L1 = new double[location_count,K];
// double[,] CA1 = new double[category_count,K];
// Dictionary<int,int> idMapperCategories = new Dictionary<int,int>();
// Dictionary<int,int> idMapperLocations =new Dictionary<int,int>();
// Dictionary<int,int> idMapperUser = new Dictionary<int,int>();
//
//
// initMatrixNormal (sub_all_data.AllUsers, ref U3, ref idMapperUser, K);
// initMatrixNormal (sub_all_data.AllUsers, ref U4, ref idMapperUser, K);
// initMatrixNormal (sub_all_data.AllUsers, ref U1, ref idMapperUser, K);
// initMatrixNormal (sub_all_data.AllUsers, ref U2, ref idMapperUser, K);
// initMatrixNormal(sub_all_data.AllItems, ref L1,ref idMapperLocations, K);
// initMatrixNormal(venueCategoryMapper.Values.ToList ().Distinct ().ToList(), ref CA1,ref idMapperCategories ,K);
// Console.WriteLine (idMapperCategories.Keys.ToList ().Count);
// for (int i = 0; i < idMapperCategories.Keys.ToList ().Count; i++) {
// Console.Write (idMapperCategories.Keys.ToList () [i]);
// Console.Write (", ");
// Console.Write (idMapperCategories.Values.ToList () [i]);
// }
// List<MyMediaLite.Eval.ItemRecommendationEvaluationResults> result_list = new List<MyMediaLite.Eval.ItemRecommendationEvaluationResults> ();
//
// double mu = 0.1;
// double best_mu = 0;
// double best_mu_value = 0;
// while (mu <= 1+0.01) { // +0.01 because of double rounding error
// Console.WriteLine ("Start with mu = " + mu.ToString ());
// WeatherItemRecommender recommender = new WeatherItemRecommender (U1, U2, U3, U4, CA1, L1,idMapperLocations, idMapperCategories, idMapperUser);
// recommender.connection_string = connection;
// recommender.Ratings = training_data;
// recommender.Validation = validation_data;
// recommender.weather_category_aware = true;
// recommender.weather_aware = false;
// recommender.max_iter = 250;
// recommender.evaluation_at = 251;
// recommender.beta = 0f;
// recommender.mu = mu;
// DateTime start_time = DateTime.Now;
// recommender.Train ();
// Console.Write ("Total Training time needed:");
// Console.WriteLine (((TimeSpan)(DateTime.Now - start_time)).TotalMilliseconds);
// Console.WriteLine ("Final results in this iteration:");
// MyMediaLite.Eval.ItemRecommendationEvaluationResults results = MyMediaLite.Eval.ItemsWeatherItemRecommender.EvaluateTime (recommender, test_data, training_data);
// foreach (var key in results.Keys)
// Console.WriteLine ("{0}={1}", key, results [key]);
// if (results ["prec@5"] > best_mu_value) {
// best_mu = mu;
// best_mu_value = results ["prec@5"];
// }
// Console.WriteLine ("Finished mu = " + mu.ToString ());
// mu += 0.1;
//
// }
// return best_mu;
// }
public static void totalTest(string data_file, int mode, int city, int iterations, string feature)
{
ITimedRatings validation_data = new TimedRatings(); // 10%
ITimedRatings test_data = new TimedRatings(); // 20%
ITimedRatings training_data = new TimedRatings(); // 70%
ITimedRatings all_data = readData(data_file);
if (mode == 1)
{
// Console.WriteLine ("Start beta tuning");
// double beta = 0.1;
// for (int i = 0; i < 5; i++) {
// Console.WriteLine ("Start beta total iteration " + i.ToString ());
// beta = startBetaTuning (all_data);
// Console.WriteLine ("Finished beta total iteration " + i.ToString ());
// Console.WriteLine ("Best beta = " + beta.ToString ());
// }
// Console.WriteLine ("End beta tuning");
}
if (mode == 2)
{
double beta = 0.2;
Console.WriteLine("Start geo base-line algo");
startIterationTestNew(all_data, false, beta, 0f, city, iterations);
Console.WriteLine("End geo base-line algo");
}
if (mode == 3)
{
double beta = 0.2;
Console.WriteLine("Start geo weather aware algo");
startIterationTest(all_data, true, beta, 0f, city, iterations);
Console.WriteLine("End geo base-line algo");
}
if (mode == 4)
{
double beta = 0.2;
Console.WriteLine("Start weather context aware algo");
startIterationTestNew(all_data, true, beta, 0f, city, iterations, feature);
Console.WriteLine("End weather context aware algo");
}
if (mode == 5)
{
Console.WriteLine("Start most popular algo");
startMostPopular(all_data);
Console.WriteLine("End most popular algo");
}
if (mode == 6)
{
Console.WriteLine("Start ItemKNN algo");
startItemKNN(data_file);
Console.WriteLine("End ItemKNN algo");
}
if (mode == 7)
{
Console.WriteLine("Start UserKNN algo");
startUserKNN(data_file);
Console.WriteLine("End UserKNN algo");
}
if (mode == 8)
{
Console.WriteLine("Start WRMF algo");
startWRMF(all_data);
Console.WriteLine("End WRMF algo");
}
if (mode == 9)
{
Console.WriteLine("Start BPRMF algo");
startBPRMF(all_data);
Console.WriteLine("End BPRMF algo");
}
}
