public string GetDetail(HitomiAnalysisArtist user)
{
Dictionary <string, double> tags_rate = new Dictionary <string, double>();
foreach (var artist in rate)
{
if (user.IsExsit(artist.Key))
{
if (tags_rate.ContainsKey(artist.Key))
{
tags_rate[artist.Key] += user.GetRate(artist.Key) * artist.Value;
}
else
{
tags_rate.Add(artist.Key, user.GetRate(artist.Key) * artist.Value);
}
}
}
var list = tags_rate.ToList();
list.Sort((p1, p2) => p2.Value.CompareTo(p1.Value));
StringBuilder builder = new StringBuilder();
foreach (var pair in list)
{
builder.Append($"{pair.Key}\r\n");
}
//builder.Append($"{pair.Key}({pair.Value}),");
return(builder.ToString());
}
public void Update()
{
HitomiAnalysisArtist user;
user = new HitomiAnalysisArtist(CustomAnalysis);
///////////////////////////////
Dictionary <string, Tuple <double, HitomiAnalysisArtist> > score = new Dictionary <string, Tuple <double, HitomiAnalysisArtist> >();
bool rms = Settings.Instance.HitomiAnalysis.UsingRMSAanlysis;
bool cos = Settings.Instance.HitomiAnalysis.UsingCosineAnalysis;
if (!rms && !cos)
{
cos = true;
}
if (rms && cos)
{
System.Windows.Forms.MessageBox.Show("[작가 추천 설정 오류] rms, cos는 동시에 사용할 수 없습니다.", "Hitomi Copy", System.Windows.Forms.MessageBoxButtons.OK, System.Windows.Forms.MessageBoxIcon.Error);
return;
}
///////////////////////////////
if (rms == true && user.GetDictionary().Count != 0)
{
double x_mean = user.GetDictionary().Sum(x => x.Value) / user.GetDictionary().Count;
double va = user.GetDictionary().Sum(x => (x.Value - x_mean) * (x.Value - x_mean));
foreach (var data in HitomiAnalysis.Instance.datas)
{
if (data.GetDictionary().Count == 0)
{
continue;
}
double y_mean = data.GetDictionary().Sum(x => x.Value) / data.GetDictionary().Count;
double a = 0.0;
HashSet <string> map = new HashSet <string>();
// 기본 형태는 rms 기반이나, rms는 예측을 하는 방법이지, 유사도 측정 방법이 아님
// 따라서 계산에 사용되지 않는 태그들은 비율에 따라 반전시켜 계산하면
// tangent가 요동치므로 신뢰성있는 결과를 얻을 수 있을 것 같음 ㅇㅅㅇ
foreach (var pair in user.GetDictionary())
{
if (data.IsExsit(pair.Key))
{
a += (data.GetRate(pair.Key) - y_mean) * (pair.Value - x_mean);
}
else
{
a += -y_mean * (pair.Value - x_mean) * (pair.Value - x_mean);
}
map.Add(pair.Key);
}
foreach (var pair in data.GetDictionary())
{
if (!map.Contains(pair.Key))
{
a += -x_mean *x_mean *Math.Abs(data.GetRate(pair.Key) - y_mean);
}
}
if (a / va >= 1.0)
{
score.Add(data.Aritst, new Tuple <double, HitomiAnalysisArtist>(va / a * 100, data));
}
else if (a / va > 0.0)
{
score.Add(data.Aritst, new Tuple <double, HitomiAnalysisArtist>(a / va * 100, data));
}
else
{
double v = -a / va;
if (v >= 1.0)
{
score.Add(data.Aritst, new Tuple <double, HitomiAnalysisArtist>(va / a * 100, data));
}
else
{
score.Add(data.Aritst, new Tuple <double, HitomiAnalysisArtist>(a / va * 100, data));
}
}
}
}
else if (cos == true && user.GetDictionary().Count != 0)
{
double s_user = Math.Sqrt(user.GetDictionary().Sum(x => x.Value * x.Value));
foreach (var data in HitomiAnalysis.Instance.datas)
{
double s_data = Math.Sqrt(data.GetDictionary().Sum(x => x.Value * x.Value));
double dist = 0.0;
if (s_user * s_data == 0.0)
{
continue;
}
foreach (var pair in user.GetDictionary())
{
if (data.IsExsit(pair.Key))
{
dist += data.GetRate(pair.Key) * pair.Value;
}
}
score.Add(data.Aritst, new Tuple <double, HitomiAnalysisArtist>(dist / (s_user * s_data) * 100, data));
}
}
///////////////////////////////
var list = score.ToList();
list.Sort((p1, p2) => p2.Value.Item1.CompareTo(p1.Value.Item1));
if (Rank != null)
{
Rank.Clear();
}
else
{
Rank = new List <Tuple <string, double, string> >();
}
foreach (var item in list)
{
Rank.Add(new Tuple <string, double, string>(item.Key, item.Value.Item1, item.Value.Item2.GetDetail(user)));
}
}
public void Update()
{
HitomiAnalysisArtist user;
if (!UserDefined)
{
user = new HitomiAnalysisArtist(HitomiLog.Instance.GetEnumerator());
}
else
{
user = new HitomiAnalysisArtist(CustomAnalysis);
}
///////////////////////////////
Dictionary <string, Tuple <double, HitomiAnalysisArtist> > score = new Dictionary <string, Tuple <double, HitomiAnalysisArtist> >();
bool xi = Settings.Instance.HitomiAnalysis.UsingXiAanlysis;
bool rms = Settings.Instance.HitomiAnalysis.UsingRMSAanlysis;
bool cos = Settings.Instance.HitomiAnalysis.UsingCosineAnalysis;
if ((!(xi ^ rms ^ cos) && (xi | rms | cos)) || (xi & rms & cos))
{
System.Windows.Forms.MessageBox.Show("[작가 추천 설정 오류] xi, rms, cos는 동시에 사용할 수 없습니다.", "Hitomi Copy", System.Windows.Forms.MessageBoxButtons.OK, System.Windows.Forms.MessageBoxIcon.Error);
return;
}
///////////////////////////////
if (!rms && !cos)
{
foreach (var pair in user.GetDictionary())
{
foreach (var data in datas)
{
if (!xi)
{
if (data.IsExsit(pair.Key))
{
if (score.ContainsKey(data.Aritst))
{
score[data.Aritst] = new Tuple <double, HitomiAnalysisArtist>(score[data.Aritst].Item1 + pair.Value * data.GetRate(pair.Key), score[data.Aritst].Item2);
}
else
{
score.Add(data.Aritst, new Tuple <double, HitomiAnalysisArtist>(pair.Value * data.GetRate(pair.Key), data));
}
}
}
else
{
if (data.IsExsit(pair.Key))
{
if (score.ContainsKey(data.Aritst))
{
score[data.Aritst] = new Tuple <double, HitomiAnalysisArtist>(score[data.Aritst].Item1 + -Math.Pow(Math.Abs(pair.Value - data.GetRate(pair.Key)), 2) * pair.Value, score[data.Aritst].Item2);
}
else
{
score.Add(data.Aritst, new Tuple <double, HitomiAnalysisArtist>(-Math.Pow(Math.Abs(pair.Value - data.GetRate(pair.Key)), 2) * pair.Value, data));
}
}
}
}
}
}
else if (rms == true && user.GetDictionary().Count != 0)
{
double x_mean = user.GetDictionary().Sum(x => x.Value) / user.GetDictionary().Count;
double va = user.GetDictionary().Sum(x => (x.Value - x_mean) * (x.Value - x_mean));
foreach (var data in datas)
{
if (data.GetDictionary().Count == 0)
{
continue;
}
double y_mean = data.GetDictionary().Sum(x => x.Value) / data.GetDictionary().Count;
double a = 0.0;
HashSet <string> map = new HashSet <string>();
// 기본 형태는 rms 기반이나, rms는 예측을 하는 방법이지, 유사도 측정 방법이 아님
// 따라서 계산에 사용되지 않는 태그들은 비율에 따라 반전시켜 계산하면
// tangent가 요동치므로 신뢰성있는 결과를 얻을 수 있을 것 같음 ㅇㅅㅇ
foreach (var pair in user.GetDictionary())
{
if (data.IsExsit(pair.Key))
{
a += (data.GetRate(pair.Key) - y_mean) * (pair.Value - x_mean);
}
else
{
a += -y_mean * (pair.Value - x_mean) * (pair.Value - x_mean);
}
map.Add(pair.Key);
}
foreach (var pair in data.GetDictionary())
{
if (!map.Contains(pair.Key))
{
a += -x_mean *x_mean *Math.Abs(data.GetRate(pair.Key) - y_mean);
}
}
if (a / va >= 1.0)
{
score.Add(data.Aritst, new Tuple <double, HitomiAnalysisArtist>(va / a * 100, data));
}
else if (a / va > 0.0)
{
score.Add(data.Aritst, new Tuple <double, HitomiAnalysisArtist>(a / va * 100, data));
}
else
{
double v = -a / va;
if (v >= 1.0)
{
score.Add(data.Aritst, new Tuple <double, HitomiAnalysisArtist>(va / a * 100, data));
}
else
{
score.Add(data.Aritst, new Tuple <double, HitomiAnalysisArtist>(a / va * 100, data));
}
}
}
}
else if (cos == true && user.GetDictionary().Count != 0)
{
double s_user = Math.Sqrt(user.GetDictionary().Sum(x => x.Value * x.Value));
foreach (var data in datas)
{
double s_data = Math.Sqrt(data.GetDictionary().Sum(x => x.Value * x.Value));
double dist = 0.0;
if (s_user * s_data == 0.0)
{
continue;
}
foreach (var pair in user.GetDictionary())
{
if (data.IsExsit(pair.Key))
{
dist += data.GetRate(pair.Key) * pair.Value;
}
}
score.Add(data.Aritst, new Tuple <double, HitomiAnalysisArtist>(dist / (s_user * s_data) * 100, data));
}
}
///////////////////////////////
// 태그 차집합 감소 연산
if (xi)
{
foreach (var pair in user.GetDictionary())
{
foreach (var data in datas)
{
if (!data.IsExsit(pair.Key) && score.ContainsKey(data.Aritst))
{
score[data.Aritst] = new Tuple <double, HitomiAnalysisArtist>(score[data.Aritst].Item1 + -pair.Value * pair.Value * pair.Value, score[data.Aritst].Item2);
}
}
}
}
///////////////////////////////
var list = score.ToList();
// 하한값을 0으로 설정
if (xi)
{
double min_score = 0.0;
foreach (var tuple in list)
{
min_score = Math.Min(min_score, tuple.Value.Item1);
}
for (int i = 0; i < list.Count; i++)
{
list[i] = new KeyValuePair <string, Tuple <double, HitomiAnalysisArtist> >(list[i].Key, new Tuple <double, HitomiAnalysisArtist>((list[i].Value.Item1 - min_score) * 100, list[i].Value.Item2));
}
}
///////////////////////////////
if (FilterArtists)
{
Dictionary <string, int> artists_galleris_count_log = new Dictionary <string, int>();
foreach (var artist in HitomiLog.Instance.GetEnumerator().Where(data => data.Artists != null).SelectMany(data => data.Artists))
{
if (artists_galleris_count_log.ContainsKey(artist))
{
artists_galleris_count_log[artist] += 1;
}
else
{
artists_galleris_count_log.Add(artist, 1);
}
}
for (int i = 0; i < list.Count; i++)
{
if (artists_galleris_count_log.ContainsKey(list[i].Key))
{
float mul = 1 - (float)artists_galleris_count_log[list[i].Key] / list[i].Value.Item2.MetadataCount;
list[i] = new KeyValuePair <string, Tuple <double, HitomiAnalysisArtist> >(list[i].Key, new Tuple <double, HitomiAnalysisArtist>(list[i].Value.Item1 * mul, list[i].Value.Item2));
}
}
}
///////////////////////////////
if (MustInclude)
{
int count = CustomAnalysis.Count;
for (int i = 0; i < list.Count; i++)
{
if (CustomAnalysis.Any(x => !list[i].Value.Item2.IsExsit(x.Item1)))
{
list.RemoveAt(i--);
}
}
}
///////////////////////////////
list.Sort((p1, p2) => p2.Value.Item1.CompareTo(p1.Value.Item1));
if (Rank != null)
{
Rank.Clear();
}
else
{
Rank = new List <Tuple <string, double, string> >();
}
foreach (var item in list)
{
Rank.Add(new Tuple <string, double, string>(item.Key, item.Value.Item1, item.Value.Item2.GetDetail(user)));
}
}