forked from amrk000/ML-Recommender-System
-
Notifications
You must be signed in to change notification settings - Fork 0
/
MLManager.cs
267 lines (199 loc) · 10.6 KB
/
MLManager.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
258
259
260
261
262
263
264
265
266
267
using System;
using System.Collections.Generic;
using System.Linq;
using System.Threading.Tasks;
using System.IO;
using Microsoft.ML;
using Microsoft.ML.Trainers;
using Microsoft.Extensions.ML;
using System.Runtime.CompilerServices;
using static Microsoft.ML.DataOperationsCatalog;
using RecommenderEngine.DataModels.ML;
using System.Text.Json;
namespace RecommenderEngine
{
public static class MLManager
{
private static MLContext mlContext;
private static String modelName;
private static String modelPath;
private static String lastDatasetJsonPath;
private static String lastDatasetCsvPath;
private static String lastTrainingResultsJsonPath;
private static String lastTrainingResultsStringPath;
private static String trainingResults;
private static TrainingResultsModel trainingResultsModel;
private static double datasetSplitRatio;
private static int numberOfIterations;
private static float learningRate;
private static int approximationRank;
private static bool training;
static MLManager()
{
mlContext = new MLContext(); //ML Initialization
//Setting ML Model Name & Path
modelName = "RecommenderModel";
modelPath = Path.Combine(AppDomain.CurrentDomain.BaseDirectory, "RecommenderModel.zip").ToString();
//Setting Last Training Dataset Files
lastDatasetJsonPath = Path.Combine(AppDomain.CurrentDomain.BaseDirectory, "lastDatasetBackup.json").ToString();
lastDatasetCsvPath = Path.Combine(AppDomain.CurrentDomain.BaseDirectory, "lastDatasetBackup.csv").ToString();
//Setting Last Training Results Files
lastTrainingResultsJsonPath = Path.Combine(AppDomain.CurrentDomain.BaseDirectory, "lastTrainingResults.json").ToString();
lastTrainingResultsStringPath = Path.Combine(AppDomain.CurrentDomain.BaseDirectory, "lastTrainingResults.txt").ToString();
trainingResults = "";
trainingResultsModel = new TrainingResultsModel();
//ML Training Algorithm Configs
datasetSplitRatio = 0.20; //Split Ratio of Dataset (Train / Test) 0.20 or 0.30 is recommended
numberOfIterations = 5000; //The Total training iterations count
learningRate = 0.001f; //The rate of weights adjustments
approximationRank = 10; //K-Rank of matrix factorization
training = false;
}
public static String GetModelName() {
return modelName;
}
public static String GetModelPath()
{
return modelPath;
}
public static bool DoesModelExists()
{
return File.Exists(modelPath);
}
public static bool DoesLastDatasetJsonExists()
{
return File.Exists(lastDatasetJsonPath);
}
public static String GetLastDatasetJson()
{
return File.ReadAllText(lastDatasetJsonPath);
}
public static bool DoesLastDatasetCsvExists()
{
return File.Exists(lastDatasetCsvPath);
}
public static String GetLastDatasetCsv()
{
return File.ReadAllText(lastDatasetCsvPath);
}
public static String GetLastTrainResults()
{
if (trainingResults.Length > 0) return trainingResults;
else if(File.Exists(lastTrainingResultsStringPath)) {
trainingResults = File.ReadAllText(lastTrainingResultsStringPath);
return trainingResults;
}
else return "No Previous Training Results Found";
}
public static TrainingResultsModel GetLastTrainResultsModel()
{
if (trainingResultsModel.Status == null && File.Exists(lastTrainingResultsJsonPath))
{
trainingResultsModel = JsonSerializer.Deserialize<TrainingResultsModel>(File.ReadAllText(lastTrainingResultsJsonPath));
return trainingResultsModel;
}
return trainingResultsModel;
}
public static bool IsTrainingInProgress()
{
return training;
}
public static MLRatingPrediction Predict(MLDataModel dataModel, PredictionEnginePool<MLDataModel, MLRatingPrediction> predictionEnginePool)
{
return predictionEnginePool.Predict(modelName: modelName, dataModel);
}
public static string TrainAndBuildModel(MLDataModel[] DataSet)
{
training = true;
try
{
double startTime = DateTime.UtcNow.TimeOfDay.TotalSeconds;
trainingResults = "Model Training Request | " + DateTime.UtcNow.ToString() + " (UTC)\n";
trainingResultsModel = new TrainingResultsModel();
trainingResultsModel.StartTime = DateTime.UtcNow.ToString() + " (UTC)";
trainingResultsModel.Status = TrainingResultsModel.STATUS_TRAINING;
trainingResultsModel.Details = "Training is In Progress";
IDataView dataView = LoadData(DataSet);
TrainTestData trainTestData = mlContext.Data.TrainTestSplit(dataView, testFraction: datasetSplitRatio);
ITransformer model = BuildAndTrainModel(trainTestData.TrainSet);
EvaluateModel(trainTestData.TestSet, model);
SaveModel(trainTestData.TrainSet.Schema, model);
double endTime = DateTime.UtcNow.TimeOfDay.TotalSeconds;
trainingResults += "Total Training Time: " + String.Format("{0:0.00}", (endTime - startTime)) + "s";
trainingResultsModel.TotalTime = String.Format("{0:0.00}", (endTime - startTime)) + "s";
SaveLastTrainingResults();
BackupDatasetCsv(dataView);
BackupDatasetJson(DataSet);
}
catch (Exception e) {
trainingResults += "\nModel Training Failed !";
trainingResults += "\nError: " + e.Message;
trainingResultsModel.Status = TrainingResultsModel.STATUS_ERROR;
trainingResultsModel.Details = "Model Training Failed !";
}
training = false;
return trainingResults;
}
private static IDataView LoadData(MLDataModel[] DataSet)
{
IDataView dataView = mlContext.Data.LoadFromEnumerable<MLDataModel>(DataSet);
long dataRowsCount = dataView.GetRowCount().Value;
trainingResults += "\nDataSet ("+ dataRowsCount + " Records | Split Ratio: "+ datasetSplitRatio + "): "+ "\nTrainig Data Records: " + dataRowsCount * (1-datasetSplitRatio) + "\nTesting Data Records: " + dataRowsCount*datasetSplitRatio + "\n";
trainingResultsModel.DataSet.RecordsCount = dataRowsCount;
trainingResultsModel.DataSet.SplitRatio = datasetSplitRatio;
trainingResultsModel.DataSet.TrainRecordsCount = (long)(dataRowsCount * (1 - datasetSplitRatio));
trainingResultsModel.DataSet.TestRecordsCount = (long)(dataRowsCount * datasetSplitRatio);
return dataView;
}
private static ITransformer BuildAndTrainModel(IDataView trainingDataView)
{
IEstimator<ITransformer> estimator = mlContext.Transforms.Conversion.MapValueToKey(outputColumnName: "userIdEncoded", inputColumnName: "userId")
.Append(mlContext.Transforms.Conversion.MapValueToKey(outputColumnName: "itemIdEncoded", inputColumnName: "itemId"));
var options = new MatrixFactorizationTrainer.Options
{
MatrixColumnIndexColumnName = "userIdEncoded",
MatrixRowIndexColumnName = "itemIdEncoded",
LabelColumnName = "Label",
NumberOfIterations = numberOfIterations, //default 20
LearningRate = learningRate, //default 0.1
ApproximationRank = approximationRank //default 8
};
trainingResults += "\nTraining Configs:\nNumber Of Iterations: " + numberOfIterations + "\nLearning Rate: " + learningRate + "\nApproximation Rank: " + approximationRank + "\n";
trainingResultsModel.TrainingConfigs.NumberOfIterations = numberOfIterations;
trainingResultsModel.TrainingConfigs.LearningRate = learningRate;
trainingResultsModel.TrainingConfigs.ApproximationRank = approximationRank;
var trainerEstimator = estimator.Append(mlContext.Recommendation().Trainers.MatrixFactorization(options));
ITransformer model = trainerEstimator.Fit(trainingDataView);
return model;
}
private static void EvaluateModel(IDataView testDataView, ITransformer model)
{
trainingResults += "\nTraining Metrices: \n";
var prediction = model.Transform(testDataView);
var trainingMetrics = mlContext.Regression.Evaluate(prediction, labelColumnName: "Label", scoreColumnName: "Score");
trainingResults += "RMSE: " + trainingMetrics.RootMeanSquaredError.ToString() + " (Lower is Better)\n";
trainingResults += "RSquared: " + trainingMetrics.RSquared.ToString() + " (Higher is Better , Max: 1.0)\n";
trainingResultsModel.TrainingMetrices.Rmse = trainingMetrics.RootMeanSquaredError;
trainingResultsModel.TrainingMetrices.RSquared = trainingMetrics.RSquared;
}
private static void SaveModel(DataViewSchema trainingDataViewSchema, ITransformer model)
{
mlContext.Model.Save(model, trainingDataViewSchema, modelPath);
trainingResults += "\nModel is Trained & Built Successfully\n";
trainingResultsModel.Status = TrainingResultsModel.STATUS_COMPLETED;
trainingResultsModel.Details = "Model is Trained & Built Successfully";
}
private static void BackupDatasetJson(MLDataModel[] DataSet)
{
File.WriteAllText(lastDatasetJsonPath, JsonSerializer.Serialize(DataSet));
}
private static void BackupDatasetCsv(IDataView dataset)
{
mlContext.Data.SaveAsText(dataset, File.CreateText(lastDatasetCsvPath).BaseStream, separatorChar: ',', headerRow: true, schema: false);
}
private static void SaveLastTrainingResults() {
File.WriteAllText(lastTrainingResultsStringPath, trainingResults);
File.WriteAllText(lastTrainingResultsJsonPath, JsonSerializer.Serialize(trainingResultsModel));
}
}
}