void FillTweetInfo()
{
Analyser analyser = new Analyser(tweetDatas, labels);
analyser.Analyse1();
tweetInfo.Text = "";
if (analyser.LabelFreq.Count == 0)
{
return;
}
for (int i = 0; i < labels.Length; i++)
{
if (analyser.LabelFreq.ContainsKey(labels[i]) == true)
{
tweetInfo.Text += labels[i] + ": " + analyser.LabelFreq[labels[i]] + "\n";
}
else
{
tweetInfo.Text += labels[i] + ": 0" + "\n";
}
}
tweetInfo.Text += "\nTotal: " + analyser.Tweets.Length + "\n";
}
void SetAnalyser(Analyser _analyser)
{
trainingTweets = _analyser.TrainingTweets;
testingTweets = _analyser.TestingTweets;
labels = _analyser.Labels;
wordDict = _analyser.WordDictionary;
labelDict = _analyser.LabelDictioanary;
}
public Trainer(Analyser _analyser, int _hiddenNodes, double _learningRate, TrainingType type)
{
analyser = _analyser;
if (analyser != null)
{
trainingTweets = _analyser.TrainingTweets;
testingTweets = _analyser.TestingTweets;
labels = _analyser.Labels;
wordDict = _analyser.WordDictionary;
labelDict = _analyser.LabelDictioanary;
nn = new NeuralNetwork(
(type == TrainingType.WORD_TRAINING) ?_analyser.MaxWordPerTweet : 4,
_hiddenNodes,
_analyser.LabelCount,
_learningRate
);
}
}
private void Test(TrainingType type, Analyser analyser = null)
{
if (string.IsNullOrEmpty(filePath))
{
MessageBox.Show("you have to choose a file to train....");
return;
}
if (analyser == null)
{
analyser = trainer.Analyser;
}
TweetData[] twData = analyser.TestingTweets;
progressBar.Value = 0;
progressBar.Maximum = twData.Length;
List <List <Tuple <int, double> > > list;
if (type == TrainingType.WORD_TRAINING)
{
list = trainer.Test1(progressBar);
}
else
{
list = trainer.Test2(progressBar);
}
confussion = new int[labels.Length, labels.Length];
richtxtAnnResult.Text = "";
for (int i = 0; i < analyser.TestingTweets.Length; i++)
{
richtxtAnnResult.AppendText("TWEET:\n\n");
richtxtAnnResult.AppendText(analyser.TestingTweets[i].tweet + "\n\n");
double max = double.MinValue;
int index = 0;
richtxtAnnResult.AppendText("PREDICTION:\n\n");
for (int j = 0; j < list[i].Count; j++)
{
// We may want to percentage of prediction
string val = labels[list[i][j].Item1];
double percentage = list[i][j].Item2 * 100;
if (percentage > max)
{
max = percentage;
index = list[i][j].Item1;
}
richtxtAnnResult.AppendText(val + ": " + percentage.ToString("F2") + "%\n");
}
int correct_Index = Array.IndexOf(labels, analyser.TestingTweets[i].users[0].labels[0]);
confussion[correct_Index, index]++;
richtxtAnnResult.AppendText("\nANSWER:\t" + analyser.TestingTweets[i].users[0].labels[0]);
richtxtAnnResult.AppendText("\n\n");
richtxtAnnResult.AppendText("######################################################n\n");
}
ShowAnalyserInfo(trainer);
}
private void TrainTest(TrainingType type)
{
if (string.IsNullOrEmpty(filePath))
{
MessageBox.Show("you have to choose a file to train....");
return;
}
btnTest.Enabled = true;
progressBar.Value = 0;
TweetData[] twData = JsonFileController.ReadDataFromJsonFile <TweetData[]>(filePath);
ParseTweets(ref twData, type);
Analyser analyser = new Analyser(twData, labels, tbTestCount.Value, chckPickRandomly.Checked);
if (type == TrainingType.WORD_TRAINING)
{
analyser.Analyse1();
}
else
{
analyser.Analyse2();
}
progressBar.Maximum = twData.Length;
trainer = new Trainer(analyser, tbHiddenNeuronCount.Value, double.Parse(lblLearningRate.Text), type);
List <List <Tuple <int, double> > > list;
if (type == TrainingType.WORD_TRAINING)
{
trainer.Train1(progressBar);
list = trainer.Test1(progressBar);
}
else
{
trainer.Train2(progressBar);
list = trainer.Test2(progressBar);
}
Test(type, analyser);
//richtxtAnnResult.Text = "";
//for (int i = 0; i < analyser.TestingTweets.Length; i++)
//{
// richtxtAnnResult.AppendText("TWEET:\n\n");
// richtxtAnnResult.AppendText(analyser.TestingTweets[i].tweet + "\n\n");
// richtxtAnnResult.AppendText("PREDICTION:\n\n");
// for (int j = 0; j < list[i].Count; j++)
// {
// string val = labels[list[i][j].Item1];
// double percentage = list[i][j].Item2 * 100;
// richtxtAnnResult.AppendText(val + ":\t" + percentage.ToString("F2") + "%\n");
// }
// richtxtAnnResult.AppendText("\nANSWER: " + analyser.TestingTweets[i].users[0].labels[0]);
// richtxtAnnResult.AppendText("\n\n");
// richtxtAnnResult.AppendText("#################################################################\n\n");
//}
ShowAnalyserInfo(trainer);
}
