private void button2_Click(object sender, EventArgs e)
{
try
{
//路径
var fullPath = Path.GetFullPath(textBox2.Text);
var dir = Path.GetDirectoryName(fullPath);
var trainDoc = FileUtils.RelativePath(Application.StartupPath, fullPath).Replace('\\', '/');
var finalData = Path.Combine(dir, "evaluate.txt");
File.Delete(finalData);
//初始参数
int cnt = Convert.ToInt32(txtCnt.Text);
double init = Convert.ToDouble(txtIni.Text);
double inc = Convert.ToDouble(txtInc.Text);
//初始化数据表
InitTable();
//开始评估
for (int i = 0; i < cnt; i++)
{
if (rbK.Checked)
{
args = string.Format("-est -dir -beta {0} -ntopics {1} -niters 1000 -savestep 500 -dfile {2}", 0.1, (int)(init + i * inc), trainDoc);
}
else
{
args = string.Format("-est -beta {0} -ntopics {1} -niters 1000 -savestep 500 -dfile {2}", init + i * inc, 12, trainDoc);
}
ProcessUtils.StartProcess("lda.exe", args);
var tt = FileUtils.ReadAsList(Path.Combine(dir, "model-final.theta"));
var pp = FileUtils.ReadAsList(Path.Combine(dir, "model-final.phi"));
var aa = FileUtils.ReadAsList(Path.Combine(dir, "model-final.tassign"));
var d = LDA.GetPerplexity(pp, tt, aa);
table.Rows.Add(init + i * inc, d);
label6.Text = string.Format("已完成:{0}/{1}", i + 1, cnt);
}
//设置图表
chartControl1.Series[0].DataSource = table;
chartControl1.Series[0].SetDataMembers("x", "y");
var xyDiagram = (DevExpress.XtraCharts.XYDiagram)(chartControl1.Diagram);
xyDiagram.AxisX.Title.Text = rbK.Checked ? "主题个数K" : "超参数β";
xyDiagram.EnableAxisYScrolling = true;
xyDiagram.EnableAxisYZooming = true;
chartControl1.Titles[0].Text = rbK.Checked ? "主题个数K对聚类的影响" : "超参数β对聚类的影响";
//保存模型评估数据
var sw = File.AppendText(finalData);
for (int i = 0; i < table.Rows.Count; i++)
{
sw.WriteLine(table.Rows[i]["x"] + "\t" + table.Rows[i]["y"]);
}
sw.Close();
}
catch (Exception ex)
{
MessageBox.Show(ex.Message);
//throw;
}
}
private void button2_Click(object sender, EventArgs e)
{
dir = Path.GetDirectoryName(Path.GetFullPath(textBox2.Text));
StringBuilder sb = new StringBuilder();
sb.Append("-est ");
if (txtB.Text != "")
{
sb.AppendFormat(" -beta {0} ", txtB.Text);
}
if (txtK.Text != "")
{
sb.AppendFormat(" -ntopics {0} ", txtK.Text);
}
if (txtI.Text != "")
{
sb.AppendFormat(" -niters {0} ", txtI.Text);
}
if (txtA.Text != "")
{
sb.AppendFormat(" -alpha {0} ", txtA.Text);
}
sb.AppendFormat(" -savestep 300 -twords 15 -dfile {0} ", textBox2.Text);
ProcessUtils.StartProcess("lda.exe", sb.ToString());
var data = LDA.ReadTopicWordsWeight(Path.Combine(dir, "model-final.twords"));
ShowWords(data);
MessageBox.Show("模型训练完成，接下来请根据高频词填写对应的主题标签，并记得保存");
}
private void button3_Click(object sender, EventArgs e)
{
try
{
string dir = Path.GetDirectoryName(Path.GetFullPath(textBox1.Text));
string testPath = "ldaTestDoc.txt";
string model = Path.GetFileNameWithoutExtension(textBox1.Text);
LDA.WriteLdaDocToFile(Preprocessor.SegmentRemoveStopWords(new string[] { textBox2.Text }.ToList(), false), Path.Combine(dir, testPath));
string args = string.Format("-inf -dir {0} -model {1} -dfile {2} -niters 20", dir, model, testPath);
ProcessUtils.StartProcess("lda.exe", args);
//控制台输出预测结果
var docTopicArray = LDA.ReadDocumentTopicData(Path.Combine(dir, "ldaTestDoc.txt.theta"));//把预测结果读入内存
Console.WriteLine("[{0}]读取主题映射...", DateTime.Now);
var topicLabel = LDA.GetTopicLabel(Path.Combine(dir, "ldaTopicLabel.txt"));//这个话题标签在执行本句代码之前需要人工指定好,根据model-final.twords文件一一对照打上标签
Console.WriteLine("[{0}]开始预测主题...", DateTime.Now);
var topicIndex = LDA.GetDocmentTopicIndexList(docTopicArray);//根据预测结果挑选出每篇文档最可能的主题的索引
label3.Text = topicLabel[topicIndex[0]];
}
catch (Exception ex)
{
MessageBox.Show(ex.Message);
//throw;
}
}
private void button1_Click(object sender, EventArgs e)
{
var docs = LDA.GetLDATrainDoc(Convert.ToInt32(textBox1.Text));
var bow = Preprocessor.SegmentRemoveStopWords(docs);
LDA.WriteLdaDocToFile(bow, textBox2.Text);
MessageBox.Show("已经生成验证集，接下来可以验证了");
}
private void button4_Click(object sender, EventArgs e)
{
string[] labels = new string[dataGridView1.Rows.Count];
for (int i = 0; i < dataGridView1.Rows.Count; i++)
{
labels[i] = dataGridView1[2, i].Value.ToString();
}
LDA.SetTopicLabel(labels, Path.Combine(dir, "ldaTopicLabel.txt"));
MessageBox.Show("模型保存成功");
}
private void button2_Click(object sender, EventArgs e)
{
try
{
var fullPath = Path.GetFullPath(textBox1.Text);
var dir = Path.GetDirectoryName(fullPath);
var inferDocFile = "ldaInferDoc.txt";
var thetaFile = inferDocFile + ".theta";
var topicLabelFile = "ldaTopicLabel.txt";
var inferDocAbsPath = Path.Combine(dir, inferDocFile);
var thetaAbsPath = Path.Combine(dir, thetaFile);
var topicLabelAbsPath = Path.Combine(dir, topicLabelFile);
Console.WriteLine("[{0}]开始预测主题...", DateTime.Now);
List <string> tidList = null;
Console.WriteLine("[{0}]获取微博...", DateTime.Now);
var inferOriginalDocs = LDA.GetLDAInferDoc(out tidList);
Console.WriteLine("[{0}]预处理...", DateTime.Now);
var inferDocs = Preprocessor.SegmentRemoveStopWords(inferOriginalDocs, false);
Console.WriteLine("[{0}]写入文件...", DateTime.Now);
LDA.WriteLdaDocToFile(inferDocs, inferDocAbsPath);
Console.WriteLine("[{0}]吉布斯采样...", DateTime.Now);
string args = string.Format("-inf -dir {0} -model {1} -dfile {2} -niters 20", dir, "model-final", inferDocFile);
ProcessUtils.StartProcess("lda.exe", args);
//控制台输出预测结果
Console.WriteLine("[{0}]读取预测结果...", DateTime.Now);
var docTopicArray = LDA.ReadDocumentTopicData(thetaAbsPath);//把预测结果读入内存
Console.WriteLine("[{0}]读取主题映射...", DateTime.Now);
var topicLabel = LDA.GetTopicLabel(topicLabelAbsPath);//这个话题标签在执行本句代码之前需要人工指定好,根据model-final.twords文件一一对照打上标签
Console.WriteLine("[{0}]开始预测主题...", DateTime.Now);
var topicIndex = LDA.GetDocmentTopicIndexList(docTopicArray);//根据预测结果挑选出每篇文档最可能的主题的索引
Console.WriteLine("[{0}]以下是微博的主题预测情况：", DateTime.Now);
LDA.PrintTopics(inferOriginalDocs, topicIndex, topicLabel);
//保存数据库
Console.WriteLine("[{0}]开始存储主题...", DateTime.Now);
LDA.SaveTopicToDB(topicIndex, topicLabel, tidList);
MessageBox.Show("任务已完成！", "提示", MessageBoxButtons.OK, MessageBoxIcon.Information);
}
catch (Exception ex)
{
MessageBox.Show(ex.Message, "出错啦", MessageBoxButtons.OK, MessageBoxIcon.Error);
}
}
