/// <summary>
/// 余弦相似度
/// </summary>
/// <param name="itemName">商品名</param>
/// <param name="Corpus">语料</param>
/// <returns></returns>
public static double CosSimilarity(string itemName, string Corpus)
{
double Ew2 = 0, Eb2 = 0, Ewb = 0;
double cosSimilarity = 0;
string splitword = SplitWordHelper.RemoveSplitWord(SplitWordHelper.SplitWords(itemName));
string[] swarray = splitword.Split(' ').Where(s => !string.IsNullOrEmpty(s)).ToArray();
Dictionary <string, double> item = new Dictionary <string, double>();
TFIDFJson tfidf = JsonConvert.DeserializeObject <TFIDFJson>(Corpus);
List <KeyValuePair <string, double> > corpus = tfidf.Data;
foreach (string sw in swarray)
{
if (!string.IsNullOrEmpty(sw))
{
string temp = splitword.Replace(" ", "");
MatchCollection m = Regex.Matches(temp, sw);
if (!item.ContainsKey(sw))
{
double tf = (double)m.Count / swarray.Length;
item.Add(sw, tf);
}
}
}
for (int i = 0; i < item.Count; i++)
{
Ew2 = Ew2 + Math.Pow(item.ElementAt(i).Value, 2);
for (int j = 0; j < corpus.Count; j++)
{
if (i == 0)
{
Eb2 = Eb2 + Math.Pow(corpus.ElementAt(j).Value, 2);
}
if (item.ElementAt(i).Key.Equals(corpus.ElementAt(j).Key))
{
Ewb = Ewb + (item.ElementAt(i).Value *corpus.ElementAt(j).Value);
}
else
{
continue;
}
}
}
cosSimilarity = Ewb / (Math.Sqrt(Ew2) * Math.Sqrt(Eb2));
return(cosSimilarity);
}
/// <summary>
/// 欧氏距离相似度
/// </summary>
/// <param name="itemName">商品名</param>
/// <param name="Corpus">语料</param>
/// <returns></returns>
public static double Euclidean(string itemName, string Corpus)
{
double sum = 0, same = 0;
double similarity = 0;
string splitword = SplitWordHelper.RemoveSplitWord(SplitWordHelper.SplitWords(itemName));
string[] swarray = splitword.Split(' ').Where(s => !string.IsNullOrEmpty(s)).ToArray();
Dictionary <string, double> item = new Dictionary <string, double>();
TFIDFJson tfidf = JsonConvert.DeserializeObject <TFIDFJson>(Corpus);
List <KeyValuePair <string, double> > corpus = tfidf.Data;
foreach (string sw in swarray)
{
if (!string.IsNullOrEmpty(sw))
{
string temp = splitword.Replace(" ", "");
MatchCollection m = Regex.Matches(temp, sw);
if (!item.ContainsKey(sw))
{
double tf = (double)m.Count / swarray.Length * 100000;
int input = (int)tf;
item.Add(sw, input);
}
}
}
for (int i = 0; i < item.Count; i++)
{
sum += Math.Pow(item.ElementAt(i).Value, 2);
for (int j = 0; j < corpus.Count; j++)
{
if (i == 0)
{
sum += Math.Pow(corpus.ElementAt(j).Value, 2);
}
if (item.ElementAt(i).Key.Equals(corpus.ElementAt(j).Key))
{
sum -= (Math.Pow(item.ElementAt(i).Value, 2) + Math.Pow(corpus.ElementAt(j).Value, 2));
same = Math.Pow((item.ElementAt(i).Value - corpus.ElementAt(j).Value), 2);
sum += same;
}
}
}
similarity = Math.Sqrt(sum);
return(similarity);
}
static void Main(string[] args)
{
//对训练文本进行分词处理
SplitWordHelper.GetSplitWordResultFiles();
//对分词后的文本进行特征提取
CommonHelper.TFIDF();
//计算每个商品分类的分类结果
double totalAccuracy = 0;
foreach (string fileName in TypeModel.GetTypeNameStringArray())
{
double accuracy = CommonHelper.CalculateAccuracy(fileName);
Console.WriteLine(fileName + "'s accuracy: " + accuracy);
totalAccuracy += accuracy;
}
Console.WriteLine("总准确率为" + (totalAccuracy / 16).ToString());
Console.ReadKey();
}