static void Main (string[] args)
{
string trainingPath = args[0];
string testPath = args[1];
int k_val = Convert.ToInt32(args[2]);
int simFunc = Convert.ToInt32(args[3]);
List<Document> TrainDocs = new List<Document>();
string sysOutput = (args[4]);
int docid = 0;
Stopwatch stopwatch = new Stopwatch();
stopwatch.Start();
List<String> Classes = ReadFile(trainingPath, ref docid, ref TrainDocs);
string line;
List<TestDoc> TestingDoc = new List<TestDoc>();
Dictionary<String, int> ConfusionDict = new Dictionary<string, int>();
string key;
int value, index;
using (StreamReader Sr = new StreamReader(testPath))
{
while ((line = Sr.ReadLine()) != null)
{
if (String.IsNullOrEmpty(line))
continue;
string[] words = line.Split(new string[] { " " }, StringSplitOptions.RemoveEmptyEntries);
TestDoc test = new TestDoc(words[0]);
for (int i = 1; i < words.Length; i++)
{
index = words[i].IndexOf(":");
key = words[i].Substring(0, index);
value = Convert.ToInt32(words[i].Substring(index + 1));
test.SumSquared += System.Math.Pow(value, 2);
if (test.wordCounts.ContainsKey(key))
test.wordCounts[key] += value;
else
test.wordCounts.Add(key, value);
}
if (simFunc == 1)
ProcessTrainEuclidean(TrainDocs, k_val, ref test, ref ConfusionDict);
else if (simFunc == 2)
ProcessTrainCosine(TrainDocs, k_val, ref test, ref ConfusionDict);
else
throw new Exception("invalid simfunc code");
TestingDoc.Add(test);
}
}
WriteSysOutput (sysOutput,TestingDoc, Classes,k_val);
WriteConfusionMatrix(Classes, ConfusionDict, "test", TestingDoc.Count);
stopwatch.Stop();
Console.WriteLine("Time elapsed: {0:hh\\:mm\\:ss}", stopwatch.Elapsed);
//Console.ReadLine();
}
public static void ProcessTrainEuclidean (List<Document> TrainDocs, int k_val, ref TestDoc test, ref Dictionary<String, int> ConfusionDict)
{
List<Score> ScoreList = new List<Score>();
double score;
foreach (var item in TrainDocs)
{
score = 0;
var keyList = item.Keys.Intersect(test.wordCounts.Keys).Distinct();
foreach (var word in keyList)
score += (item.wordCounts[word]*test.wordCounts[word]);
//(a-b)^2 = a^2 +b^2 -2 *a * b
score = test.SumSquared+item.SumSquared - (2 * score);
ScoreList.Add(new Score(item.classLabel, score));
}
ScoreList = ScoreList.OrderBy(s => s.ScoreValue).ToList();
string key;
for (int i = 0; i < k_val; i++)
{
key = ScoreList[i].classLabel;
if (test.ClassCount.ContainsKey(key))
test.ClassCount[key]++;
else
test.ClassCount.Add(key, 1);
}
test.ClassCount = test.ClassCount.OrderByDescending(x => x.Value).ToDictionary(x => x.Key, x => x.Value);
test.PredClass = test.ClassCount.Keys.First();
key = test.ActualClass + "_" + test.PredClass;
if (ConfusionDict.ContainsKey(key))
ConfusionDict[key]++;
else
ConfusionDict.Add(key, 1);
}
public static void ProcessTrainCosine (List<Document> TrainDocs, int k_val, ref TestDoc test, ref Dictionary<String, int> ConfusionDict)
{
List<Score> ScoreList = new List<Score>();
double score;
double V1=0, V2=0;
V2 = test.SumSquared;
V2 = System.Math.Sqrt(V2);
foreach (var item in TrainDocs)
{
V1 = item.SumSquared;
V1 = System.Math.Sqrt(V1);
score = 0;
var keyList = item.Keys.Intersect(test.wordCounts.Keys).Distinct();
foreach (var word in keyList)
{
score += (item.wordCounts[word] * test.wordCounts[word]);
}
score = (score) / (V1 * V2);
ScoreList.Add(new Score(item.classLabel, score));
}
ScoreList = ScoreList.OrderByDescending(s => s.ScoreValue).ToList();
string key;
for (int i = 0; i < k_val; i++)
{
key = ScoreList[i].classLabel;
if (test.ClassCount.ContainsKey(key))
test.ClassCount[key]++;
else
test.ClassCount.Add(key, 1);
}
test.ClassCount = test.ClassCount.OrderByDescending(x => x.Value).ToDictionary(x => x.Key, x => x.Value);
test.PredClass = test.ClassCount.Keys.First();
key = test.ActualClass + "_" + test.PredClass;
if (ConfusionDict.ContainsKey(key))
ConfusionDict[key]++;
else
ConfusionDict.Add(key, 1);
}