public void CategoryTestProperty()
{
Category c = new Category(2,"foo");
CountedCategory cc = new CountedCategory(c, 0, null);
Assert.AreEqual(c, cc.Category);
}
public void EqualsTestEqualsOperator()
{
CountedCategory o1 = new CountedCategory(
new Category(2, "foo"), 42, new Dictionary<string, int>());
CountedCategory o2 = new CountedCategory(
new Category(2, "foo"), 42, new Dictionary<string, int>());
Assert.IsTrue(o1 == o2);
}
public void EqualsTestInequality()
{
CountedCategory o1 = new CountedCategory(
new Category(2, "foo"), 42, new Dictionary<string, int>());
CountedCategory o2 = new CountedCategory(
new Category(2, "bar"), 33, new Dictionary<string, int>());
Assert.IsFalse(o1.Equals(o2));
}
public void EqualsTestGetHasCodesEquals()
{
CountedCategory o1 = new CountedCategory(new Category(2, "foo"), 42, new Dictionary<string, int>());
CountedCategory o2 = new CountedCategory(new Category(2, "foo"), 42, new Dictionary<string, int>());
Assert.AreEqual(o1.GetHashCode(), o2.GetHashCode());
}
public void DocumentCountTestProperty()
{
CountedCategory cc = new CountedCategory(null, 42, null);
Assert.AreEqual(42, cc.DocumentCount);
}
public void TermsTestProperty()
{
Dictionary<string, int> terms = new Dictionary<string, int>();
terms.Add("foo", 42);
CountedCategory cc = new CountedCategory(null, 0, terms);
Assert.AreEqual(terms, cc.Terms);
}
public void EqualsTestReferenceEquals()
{
CountedCategory o1 = new CountedCategory(
new Category(2, "foo"), 42, new Dictionary<string, int>());
CountedCategory o2 = o1;
Assert.IsTrue(o1.Equals(o2));
}
public void EqualsTestObjectToNullReferenceEqualsOperator()
{
CountedCategory o1 = new CountedCategory(new Category(2, "foo"), 42, new Dictionary<string, int>());
CountedCategory o2 = null;
Assert.IsFalse(o1 == o2);
}
public void EqualsTestObjectReferenceEquality()
{
CountedCategory o1 = new CountedCategory(
new Category(2, "foo"), 42, new Dictionary<string, int>());
Object o2 = (Object) new CountedCategory(
new Category(2, "foo"), 42, new Dictionary<string, int>());
Assert.IsTrue(o1.Equals(o2));
}
/// <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());
}
public void DetermineCategoryTestCorrectClassification()
{
List<string> terms = new List<string>()
{
"a",
"b",
"c",
"d",
"e",
"f",
"g",
"h",
"i",
"j",
"k",
"l",
"m"
};
List<CountedCategory> categories = new List<CountedCategory>();
Dictionary<string, int> cat1Terms = new Dictionary<string,int>();
cat1Terms["a"] = 80;
cat1Terms["d"] = 28;
cat1Terms["f"] = 0;
cat1Terms["k"] = 10;
CountedCategory cat1 = new CountedCategory(new Category(0, "cat1"),
100, cat1Terms);
categories.Add(cat1);
Dictionary<string, int> cat2Terms = new Dictionary<string,int>();
cat2Terms["c"] = 10;
cat2Terms["d"] = 12;
cat2Terms["h"] = 9;
cat2Terms["m"] = 11;
CountedCategory cat2 = new CountedCategory(new Category(1, "cat2"),
80, cat2Terms);
categories.Add(cat2);
Dictionary<string, int> cat3Terms = new Dictionary<string,int>();
cat3Terms["b"] = 60;
cat3Terms["i"] = 2;
cat3Terms["l"] = 4;
cat3Terms["m"] = 10;
CountedCategory cat3 = new CountedCategory(new Category(2, "cat3"),
200, cat3Terms);
categories.Add(cat3);
Category result = TermUtils.DetermineCategory(terms, categories);
// Category 1 should be the result (calculated manually).
Assert.AreEqual(cat1.Category, result);
}
