-
Notifications
You must be signed in to change notification settings - Fork 1
/
KMeans.cs
257 lines (212 loc) · 7.12 KB
/
KMeans.cs
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
using System;
using System.Collections.Concurrent;
using System.Collections.Generic;
using System.IO;
using System.Threading;
using System.Threading.Tasks;
namespace Kmeans
{
internal struct Point
{
internal double Norm { get; set; }
//internal Dictionary<ushort, double> Values;
internal ConcurrentDictionary<ushort, double> Values;
internal void CalculateNorm()
{
var norm = 0.0;
foreach (var val in Values)
{
norm += Math.Pow(val.Value, 2);
}
this.Norm = Math.Sqrt(norm);
}
}
public class KMeans
{
private const uint Size = 470758;
private readonly short _k;
private readonly double _error;
private readonly string _fileName;
private readonly ushort[] _clustering;
private readonly Point[] _dataset;
private readonly Point[] _centroids;
public KMeans(short k, double error, string fileName)
{
if (!File.Exists(fileName))
{
throw new InvalidOperationException("El archivo no existe.");
}
Console.WriteLine("El archivo existe!");
_k = k;
_error = error;
_fileName = fileName;
_dataset = new Point[Size];
_clustering = new ushort[Size];
_centroids = new Point[k];
}
public void Init()
{
LoadNetflixDataset();
GC.Collect();
GC.WaitForPendingFinalizers();
Parallel.For(0, Size, i =>
{
_dataset[i].CalculateNorm();
});
SetCentroidsRandomly();
IterateUntilConvergence();
PrintResults();
}
private void LoadNetflixDataset()
{
var file = new StreamReader(_fileName);
string line;
ushort movie=0;
var userPos = new Dictionary<uint, int>();
var i = 0;
while ((line = file.ReadLine()) != null)
{
if (line.EndsWith(':'))
{
movie = Convert.ToUInt16(line.Substring(0, line.Length - 1));
Console.WriteLine($"Leyendo datos de la película: {movie}");
}
else
{
var elements = line.Split(',');
var user = Convert.ToUInt32(elements[0]);
var rate = Convert.ToDouble(elements[1]);
int pos;
if (userPos.ContainsKey(user))
{
pos = userPos[user];
}
else
{
_dataset[i] = new Point
{
Values = new ConcurrentDictionary<ushort, double>()
//Values = new ConcurrentDictionary<ushort, double>()
};
userPos[user] = i;
pos = i;
i++;
}
_dataset[pos].Values[movie] = rate;
}
}
file.Close();
Console.WriteLine($"El dataset contiene {_dataset.Length} elementos \n\n");
}
private void SetCentroidsRandomly()
{
var rd = new Random();
for (var i = 0; i < _k; i++)
{
var index = rd.Next(_dataset.Length);
_centroids[i] = _dataset[index];
}
}
private double Clustering()
{
var ssd = 0.0;
Parallel.For(0, Size, i =>
{
ushort centroid;
double distance;
(centroid, distance) = ClosestCentroidTo(_dataset[i]);
_clustering[i] = centroid;
ssd += distance;
});
return ssd;
}
private (ushort, double) ClosestCentroidTo(Point point)
{
var distance = double.MaxValue;
ushort ci = 0; // Centroid index
for (ushort i = 0; i < _centroids.Length; i++)
{
var distancePrev = GetDistance(point, _centroids[i]);
if (distancePrev < distance)
{
distance = distancePrev;
ci = i;
}
}
return (ci, distance);
}
private void NewCentroids()
{
for (var i = 0; i < _centroids.Length; i++)
{
_centroids[i].Values = new ConcurrentDictionary<ushort, double>();
//_centroids[i].Values = new ConcurrentDictionary<ushort, double>();
_centroids[i].Norm = 0.0;
}
var count = new int[_k];
Parallel.For(0, Size, i =>
{
var ci = _clustering[i];
foreach (var pair in _dataset[i].Values)
{
// if (_centroids[ci].Values.ContainsKey(pair.Key))
// {
// _centroids[ci].Values[pair.Key] += pair.Value;
// }
// else
// {
// _centroids[ci].Values[pair.Key] = pair.Value;
// }
_centroids[ci].Values.AddOrUpdate(pair.Key, pair.Value, (key, oldValue) => oldValue + pair.Value);
}
Interlocked.Increment(ref count[ci]);
});
for (int i = 0; i < _centroids.Length; i++)
{
var keys = new List<ushort>(_centroids[i].Values.Keys);
foreach (var cKey in keys)
{
_centroids[i].Values[cKey] /= count[i];
}
}
for (int i = 0; i < _centroids.Length; i++)
{
_centroids[i].CalculateNorm();
}
}
private void IterateUntilConvergence()
{
var iter = 0;
var ssd = 0.0;
double d, ssdPrev;
do
{
ssdPrev = ssd;
Console.WriteLine($"Iteration {iter}");
ssd = Clustering();
Console.WriteLine($"SSD: {ssd:N3}");
NewCentroids();
iter++;
d = Math.Abs(ssdPrev - ssd);
Console.WriteLine($"---> {d}");
} while (d > _error);
}
private void PrintResults()
{
}
private static double GetDistance(Point a, Point b)
{
var dotProduct = 0.0;
foreach (var i in a.Values)
{
if (b.Values.TryGetValue(i.Key, out var bValue))
{
dotProduct += i.Value * bValue;
}
}
var normProduct = a.Norm * b.Norm;
var thetaCosine = dotProduct / normProduct;
return Math.Asin(thetaCosine);
}
}
}