/// <summary>
/// Runs the test.
/// </summary>
public static void Test()
{
string db = "cosine_similarity_database.sdf";
// Check if the file exists.
bool dbExists = File.Exists(db);
// Set up database.
SqlCeArchivist archivist = new SqlCeArchivist(db);
archivist.Open();
// Seed only if the db did not exist.
if (!dbExists)
{
// Seed database. Change the dataDir to the correct one.
string dataDir = @"C:\Users\Andreas Petersen\Documents\GitHub\NyhedsfilterP2\System\NewsTrainer\data";
DBSeeder.SeedDatabaseWithNews(archivist, dataDir, 100);
// Seed the database with redundant news. Change the dataDir to the correct one.
dataDir = @"C:\Users\Andreas Petersen\Documents\GitHub\NyhedsfilterP2\System\NewsFilter\TestProject\cosine_similarity_test";
DBSeeder.SeedDatabaseWithNews(archivist, dataDir, 100);
Console.WriteLine();
}
// Get the redundant news.
List<NewsItem> redundantNews = archivist.GetNews(new NewsQuery()
{
CategoryId = archivist.GetCategories().
Find(n => n.Name.Equals("redundant")).Id
});
WriteCosineSimilarity(archivist, redundantNews);
archivist.Close();
}
public static void CuratorTestClassInitialize(TestContext testContext)
{
// Initialize the archivist.
Archivist = new SqlCeArchivist("testDb");
Archivist.Open();
Archivist.TruncateData();
}
public static void InterestFilterTestClassInitialize(TestContext testContext)
{
// Initialize the archivist.
Archivist = new SqlCeArchivist("testDb");
Archivist.Open();
Archivist.TruncateData();
Filter = new RedundancyFilter();
}
public static void QuantityTestClassInitialize(TestContext testContext)
{
// Set the database connection
DatabaseConnection = new SqlCeConnection(
"Data Source=" + testContext.TestRunDirectory + "\\database.sdf");
// Initialize the archivist.
Archivist = new SqlCeArchivist(DatabaseConnection);
Archivist.Open();
}
/// <summary>
/// Static constructor for <c>Client</c>.
/// </summary>
static Client()
{
string path = String.Format(@"{0}\database.sdf",
Directory.GetCurrentDirectory());
Archivist = new SqlCeArchivist(
Properties.Settings.Default.databaseConnectionString);
Archivist.Open();
// Adds CleanUp to be called on process exit.
AppDomain.CurrentDomain.ProcessExit += CleanUp;
}
static void Main(string[] args)
{
SqlCeArchivist a = new SqlCeArchivist(@"smallDatabase.sdf");
a.Open();
//NewsQuery q = new NewsQuery();
//q.Limit = 1;
//a.GetTfIdfVector(a.GetNews(q)[0]);
a.TruncateData();
string dataDir = @"C:\Users\Andreas Petersen\Documents\GitHub\NyhedsfilterP2\System\NewsTrainer\data";
DBSeeder.SeedDatabaseWithNews(a, dataDir, 1);
//string newsSourcesDir = @"C:\Users\mkaloer\Documents\GitHub\NyhedsfilterP2\System\NewsTrainer\sources.json";
//DBSeeder.SeedNewsSources(a, newsSourcesDir);
a.Close();
return;
}
/// <summary>
/// The main test method.
/// </summary>
public static void Test()
{
SqlCeArchivist archivist = new SqlCeArchivist("db.sdf");
archivist.Open();
archivist.TruncateData();
string dataDir = @"C:\Users\mads\Desktop\data";
DBSeeder.SeedDatabaseWithNews(archivist, dataDir, 500);
// Mark sports category as interesting, business as uninteresting.
List<Category> categories = archivist.GetCategories();
archivist.MarkCategoryInteresting(categories.First(p => p.Name.Equals("sports")), true);
archivist.MarkCategoryInteresting(categories.First(p => p.Name.Equals("business")), false);
_random = new Random(42);
// Generate lists of news items.
List<NewsMaterial> interestingNews = new List<NewsMaterial>();
for (int i = 0; i < 500; i++)
{
interestingNews.Add(GenerateNewsMaterial(true));
}
List<NewsMaterial> uninterestingNews = new List<NewsMaterial>();
for (int i = 0; i < 500; i++)
{
uninterestingNews.Add(GenerateNewsMaterial(false));
}
// Add news to db and save ids.
List<int> interestingNewsIds = archivist.AddNews(interestingNews);
List<int> uninterestingNewsIds = archivist.AddNews(uninterestingNews);
List<int> allNewsIds = new List<int>(interestingNewsIds);
allNewsIds.AddRange(uninterestingNewsIds);
// Get news to filter.
NewsQuery query = new NewsQuery();
query.Read = ReadStatus.Unread;
List<NewsItem> news = archivist.GetNews(query);
// Filter news.
InterestFilter filter = new InterestFilter();
news = filter.Filter(archivist, news);
int correctCount = 0;
int falseCount = 0;
// Compare ids and count number of correct values.
foreach (NewsItem item in news)
{
if (interestingNewsIds.Contains(item.Id))
{
correctCount++;
}
else if (uninterestingNewsIds.Contains(item.Id))
{
falseCount++;
}
}
Console.WriteLine("Through filter: {0}", news.Count);
Console.WriteLine("False positive count: {0}", falseCount);
// Print result in console.
Console.WriteLine("{0}/{1}={2}%", correctCount,
interestingNews.Count,
(double) correctCount / (interestingNews.Count) * 100.0);
Console.ReadLine();
}
/// <summary>
/// The main test method.
/// </summary>
public static void Test()
{
string db = "redundancy_perf.sdf";
// Check if the file exists.
bool dbExists = File.Exists(db);
// Set up database.
SqlCeArchivist archivist = new SqlCeArchivist(db);
archivist.Open();
// Seed only if the db did not exist.
if (!dbExists)
{
// Seed database. Change the dataDir to the correct one.
string dataDir = @"C:\Users\Andreas Petersen\Documents\GitHub\NyhedsfilterP2\System\NewsTrainer\data";
DBSeeder.SeedDatabaseWithNews(archivist, dataDir, 100);
// Seed the database with redundant news. Change the dataDir to the correct one.
dataDir = @"C:\Users\Andreas Petersen\Documents\GitHub\NyhedsfilterP2\System\NewsFilter\TestProject\redundancy_perf_test";
DBSeeder.SeedDatabaseWithNews(archivist, dataDir, 100);
Console.WriteLine();
}
// Get the redundant news set.
List<NewsItem> redundantNews = archivist.GetNews(new NewsQuery()
{
CategoryId = archivist.GetCategories().
Find(n => n.Name.Equals("redundant")).Id
});
// Get some news not in the redundant set. They themselves might
// include some redundant news, but shouldn't be so in relation
// to the redundant news set.
List<NewsItem> nonRedundantNews =
archivist.GetNews(new NewsQuery() { Limit = 200 });
int nonRedundantNewsCount = 100 - redundantNews.Count;
// Find news not in the redundant news.
nonRedundantNews = nonRedundantNews.FindAll(n => !redundantNews.Contains(n));
nonRedundantNews.RemoveRange(
nonRedundantNewsCount - 1,
nonRedundantNews.Count - nonRedundantNewsCount);
// Assemble all the news.
List<NewsItem> allNews = new List<NewsItem>();
foreach (NewsItem n in redundantNews)
{
allNews.Add(n);
}
foreach (NewsItem n in nonRedundantNews)
{
allNews.Add(n);
}
// Set all the news as unread.
foreach (NewsItem n in allNews)
{
archivist.SetNewsReadStatus(n, false);
}
// Each list item contains a set of redundant news items.
List<List<int>> redundantNewsIds = new List<List<int>>();
redundantNewsIds.Add(new List<int>()
{
redundantNews.Find(n => n.Title.Equals("1")).Id,
redundantNews.Find(n => n.Title.Equals("2")).Id
});
redundantNewsIds.Add(new List<int>()
{
redundantNews.Find(n => n.Title.Equals("3")).Id,
redundantNews.Find(n => n.Title.Equals("4")).Id
});
redundantNewsIds.Add(new List<int>()
{
redundantNews.Find(n => n.Title.Equals("5")).Id,
redundantNews.Find(n => n.Title.Equals("6")).Id
});
redundantNewsIds.Add(new List<int>()
{
redundantNews.Find(n => n.Title.Equals("7")).Id,
redundantNews.Find(n => n.Title.Equals("8")).Id,
redundantNews.Find(n => n.Title.Equals("9")).Id
});
redundantNewsIds.Add(new List<int>()
{
redundantNews.Find(n => n.Title.Equals("10")).Id,
redundantNews.Find(n => n.Title.Equals("11")).Id
});
redundantNewsIds.Add(new List<int>()
{
redundantNews.Find(n => n.Title.Equals("12")).Id,
redundantNews.Find(n => n.Title.Equals("13")).Id
});
// Filter the news.
RedundancyFilter filter = new RedundancyFilter();
List<NewsItem> result = filter.Filter(archivist, allNews);
// Check that the result filters redundant news.
int correctCount = 0;
int falsePositiveCount = 0;
foreach(List<int> set in redundantNewsIds)
{
// Count number of news that went through the filter.
int newsCount = 0;
foreach (int id in set)
{
if (result.Exists(p => p.Id == id))
{
newsCount++;
}
}
// Count correct count and false positive count.
correctCount += set.Count - newsCount;
falsePositiveCount += newsCount != 1 ? 1 : 0;
}
// Calculate the expected correct count.
int expectedCorrectCount = 0;
foreach (List<int> s in redundantNewsIds)
{
expectedCorrectCount += s.Count - 1;
}
// Print the results.
Console.WriteLine("Redundant news removed: {0}/{1}, false positives: {2}",
correctCount,
expectedCorrectCount,
falsePositiveCount);
archivist.Close();
}
public static void AggregatorTestInitialize(TestContext testContext)
{
// Set the database connection
DatabaseConnection = new SqlCeConnection(
"Data Source=" + testContext.TestRunDirectory + "\\database.sdf");
// Initialize the archivist.
Archivist = new SqlCeArchivist(DatabaseConnection);
Archivist.Open();
AggregatorClass = new PrivateType(typeof(Aggregator));
}
/// <summary>
/// The main test method. Call this to run the test.
/// </summary>
public static void Test()
{
SqlCeArchivist a = new SqlCeArchivist(@"C:\Users\mads\Documents\GitHub\NyhedsfilterP2\System\NewsFilter\Experiment\Database.sdf");
a.Open();
// Uncomment to seed database. This should be done on the first test run.
// Replace with dir to data.
// string dataDir = @"C:\Users\mads\Documents\GitHub\NyhedsfilterP2\System\NewsTrainer\data\";
// DBSeeder.SeedDatabaseWithNews(a, dataDir, 3000);
List<Category> categories = a.GetCategories();
// Training data listed by category.
Dictionary<int, List<NewsItem>> trainingData =
new Dictionary<int, List<NewsItem>>();
List<NewsItem> testData = new List<NewsItem>();
// Store results here:
Dictionary<int, TestResult[]> results =
new Dictionary<int, TestResult[]>();
NewsQuery query = new NewsQuery();
foreach (Category category in categories)
{
// Now find testData by setting an offset and a new limit.
query.Offset = 10 * NEWS_COUNT_INTERVAL;
query.Limit = 1000;
query.CategoryId = category.Id;
testData.AddRange(a.GetNews(query));
results.Add(category.Id, new TestResult[10]);
}
for (int i = 0; i < 10; i++)
{
List<CountedCategory> categoryDescriptions = new List<CountedCategory>();
foreach (Category category in categories)
{
// First find training data.
query.Offset = i * NEWS_COUNT_INTERVAL;
query.Limit = NEWS_COUNT_INTERVAL;
query.CategoryId = category.Id;
if (!trainingData.ContainsKey(category.Id))
{
trainingData.Add(category.Id, new List<NewsItem>());
}
// Add to list
trainingData[category.Id].AddRange(a.GetNews(query));
// Save as CountedCategory.
CountedCategory cat = new CountedCategory();
cat.DocumentCount = trainingData[category.Id].Count;
cat.Category = category;
// Add all term frequencies.
foreach (NewsItem news in trainingData[category.Id])
{
foreach (Term t in news.Terms)
{
if (!cat.Terms.ContainsKey(t.TermName))
{
cat.Terms[t.TermName] = 0;
}
cat.Terms[t.TermName] += t.Frequency;
}
}
categoryDescriptions.Add(cat);
}
// Now test classifier for each test data item.
foreach (NewsItem item in testData)
{
List<string> terms = new List<string>();
// Add all terms to list (all occurrences).
foreach (Term t in item.Terms)
{
for (int j = 0; j < t.Frequency; j++)
{
terms.Add(t.TermName);
}
}
if (results[item.Category.Id][i] == null)
{
results[item.Category.Id][i] = new TestResult(0, 0);
}
// Classify and save result.
if (TermUtils.DetermineCategory(
terms, categoryDescriptions).Id == item.Category.Id)
{
results[item.Category.Id][i].CorrectCount += 1;
}
results[item.Category.Id][i].Count += 1;
}
}
a.Close();
// Write to CSV.
StringBuilder csv = new StringBuilder();
// Write headers.
csv.Append("data-count;");
foreach(Category c in categories)
{
csv.Append(c.Name);
csv.Append(";");
}
csv.AppendLine("total");
// Iterate through each row.
for(int i = 0; i < 10; i++)
{
// Write data-count column.
csv.Append(((i + 1) * NEWS_COUNT_INTERVAL).ToString());
csv.Append(";");
int correctSum = 0;
int countSum = 0;
// Print each category.
foreach (KeyValuePair<int, TestResult[]> result in results)
{
// Calculate average + add to string.
csv.Append(
(result.Value[i].CorrectCount /
(double)result.Value[i].Count * 100.0).ToString());
csv.Append(";");
correctSum += result.Value[i].CorrectCount;
countSum += result.Value[i].Count;
}
// Write total average
csv.AppendLine(
(correctSum /
(double)countSum * 100.0).ToString());
}
// Write to CSV.
File.WriteAllText("output.csv", csv.ToString());
}