Accord C4.5决策树算法（C# C4.5决策树算法）

算法直接调用Accord.net Framework封装好的方法，所以要下载Accord机器学习的DLL，下载地址见Accord官网。C4.5算法要引用的包括：

using Accord.IO;

using Accord.MachineLearning.DecisionTrees;

using Accord.MachineLearning.DecisionTrees.Learning;

using Accord.MachineLearning.DecisionTrees.Rules;

using Accord.Math;

using Accord.Statistics.Analysis;

using Accord.Statistics.Filters;

using AForge;

using System.Numerics;

我是写成winform窗体程序形式的，实现了五个按钮的Click()事件，分别用于：

1. MenuFileOpen_Click ()：导入测试数据集；

2. upAttributeType_Click()：导入测试数据的属性类别，属性类别分为Continuous(数值型)和Symbolic（符号型）；

3. upTestingData_Click()：导入训练数据集；

4. btnCreate_Click()：建立分类模型；

5. btnTestingRun_Click()：测试模型。

算法代码：

using System;
using System.Collections.Generic;
using System.ComponentModel;
using System.Data;
using System.Drawing;
using System.Linq;
using System.Text;
using System.Threading.Tasks;
using System.Windows.Forms;
using System.IO;
using Accord.IO;
using Accord.MachineLearning.DecisionTrees;
using Accord.MachineLearning.DecisionTrees.Learning;
using Accord.MachineLearning.DecisionTrees.Rules;
using Accord.Math;
using Accord.Statistics.Analysis;
using Accord.Statistics.Filters;
using AForge;
using System.Numerics;

namespace 粗糙集属性选择_V1._0.DecisionTrees
{
public partial class DecisionTreesMainForm : DockContent
{
// 初始化树
DecisionTree tree;

//训练数据表
DataTable myDT = new DataTable();
//属性类别表
 DataTable myDTAttType = new DataTable();
//测试数据表
 DataTable myTestDT = new DataTable();

//属性数值化的矩阵
Codification codebook;

public DecisionTreesMainForm()
{
InitializeComponent();

dgvLearningSource.AutoGenerateColumns = true;
dgvPerformance.AutoGenerateColumns = false;

openFileDialog.InitialDirectory = Path.Combine(Application.StartupPath, "Resources");
}

/// <summary>
///   进行学习并创建决策树
/// </summary>
///
private void btnCreate_Click(object sender, EventArgs e)
{
if (dgvLearningSource.DataSource == null)
{
MessageBox.Show("请先导入训练数据！");
return;
}
if (attributeTypeGridView.DataSource == null) {
MessageBox.Show("请先导入数据的属性类型！");
return;
}

// 完成和给定的数据保存修改
dgvLearningSource.EndEdit();

// MessageBox.Show(myDT.Columns.Count + "");

//初始化columnNames
int myDTColNum = myDT.Columns.Count;
String[] colNM = new string[myDTColNum];

for (int i = 0; i < myDTColNum; i++) {
colNM[i] = myDT.Columns[i].ColumnName;
}

//把属性类别放入hashMap
Dictionary<String, String> attTypeMap = new Dictionary<string, string>();
for (int i = 0, attTypeSum = myDTAttType.Rows.Count;i < attTypeSum ; i++) {
attTypeMap.Add(myDTAttType.Rows[i][0].ToString(),myDTAttType.Rows[i][1].ToString());
}

// 对属性进行数值化
try
{
codebook = new Codification(myDT, colNM);
}
catch (Exception ex) {
MessageBox.Show("运行出现异常！" + ex.Message);
}

DecisionVariable[] attributes = new DecisionVariable[myDTColNum - 1];
for (int i = 0; i < myDTColNum - 1; i++)
{
String getAttType;
attTypeMap.TryGetValue(colNM[i], out getAttType);
//Console.WriteLine("----" + getAttType);
if (getAttType.Equals("Symbolic"))
{
attributes[i] = new DecisionVariable(colNM[i], codebook[colNM[i]].Symbols);
}
else {
attributes[i] = new DecisionVariable(colNM[i], DecisionVariableKind.Continuous);
}

}

int classCount = codebook[colNM[myDTColNum - 1]].Symbols;

            // 创建决策树
tree = new DecisionTree(attributes, classCount);

//创建C4.5决策树
C45Learning c45 = new C45Learning(tree);

// 进行训练学习
DataTable symbols = codebook.Apply(myDT);
String[] inputColStr = new string[myDTColNum - 1];
for (int i = 0; i < myDTColNum - 1; i++) {
inputColStr[i] = colNM[i];
// MessageBox.Show(inputColStr[i]);
}
double[][] inputs = symbols.ToArray(inputColStr);
int[] outputs = symbols.ToArray<int>(colNM[myDTColNum - 1]);

double error = c45.Run(inputs, outputs);

//MessageBox.Show(codebook.Translate("A1","yong")+"");

// 显示树结构
decisionTreeView1.TreeSource = tree;

lbStatus.Text = "树创建完成，可以点击其它选项查看结果！";

}

/// <summary>
///   测试模型.
/// </summary>
///
private void btnTestingRun_Click(object sender, EventArgs e)
{
if (tree == null)
{
MessageBox.Show("请先建立模型！");
return;
}

if (dgvTestingSource.DataSource == null) {
MessageBox.Show("请先上传测试数据！");
return;
}

int testDTRowSum = myTestDT.Rows.Count;
int testDTColSum = myTestDT.Columns.Count;

int[] expected = new int[testDTRowSum];
for (int i = 0; i < testDTRowSum; i++) {
//将测试数据的每一条记录的类别转化成决策树中的类别（类标）
 expected[i] = codebook.Translate(myTestDT.Columns[testDTColSum - 1].ColumnName, myTestDT.Rows[i][testDTColSum - 1].ToString());
//Console.WriteLine("expected-----" + expected[i]);
}

int[] actual = new int[testDTRowSum];
String[] myTestDTcolumnsName = new string[testDTColSum];
for(int i=0;i < testDTColSum;i++){
myTestDTcolumnsName[i] = myTestDT.Columns[i].ColumnName;
}

 for (int i = 0; i < testDTRowSum; i++)
            {
                try
                {
                    actual[i] = tree.Compute(codebook.Translate(myTestDT.Rows[i], myTestDTcolumnsName));
                }
                catch (Exception ex)
                {
                    MessageBox.Show("出现异常！" + ex.Message);
                }
            }

            // 计算分类性能.
            ConfusionMatrix confusionMatrix = new ConfusionMatrix(actual, expected, 1, 0);
            dgvPerformance.DataSource = new[] { confusionMatrix };
        }

        //导入数据集
        private void MenuFileOpen_Click(object sender, EventArgs e)
        {
            if (openFileDialog.ShowDialog(this) == DialogResult.OK)
            {
                string filename = openFileDialog.FileName;
                string extension = Path.GetExtension(filename);
                if (extension == ".xls" || extension == ".xlsx")
                {
                    ExcelReader db = new ExcelReader(filename, true, false);
                    TableSelectDialog t = new TableSelectDialog(db.GetWorksheetList());

                    if (t.ShowDialog(this) == DialogResult.OK)
                    {
                        DataTable tableSource = db.GetWorksheet(t.Selection);

                        this.dgvLearningSource.DataSource = tableSource;
                        myDT = tableSource.Copy();

                    }
                }
            }

            lbStatus.Text = "导入训练数据集后，点击“创建树”按钮后开始创建树！";
        }

        //导入属性类别
        private void upAttributeType_Click(object sender, EventArgs e)
        {
            if (openFileDialog.ShowDialog(this) == DialogResult.OK)
            {
                string filename = openFileDialog.FileName;
                string extension = Path.GetExtension(filename);
                if (extension == ".xls" || extension == ".xlsx")
                {
                    ExcelReader db = new ExcelReader(filename, true, false);
                    TableSelectDialog t = new TableSelectDialog(db.GetWorksheetList());

                    if (t.ShowDialog(this) == DialogResult.OK)
                    {
                        DataTable tableSource = db.GetWorksheet(t.Selection);
                        this.attributeTypeGridView.DataSource = tableSource;
                        myDTAttType = tableSource.Copy();

                    }
                }
            }

        }

        //导入测试数据集
        private void upTestingData_Click(object sender, EventArgs e)
        {
            if (openFileDialog.ShowDialog(this) == DialogResult.OK)
            {
                string filename = openFileDialog.FileName;
                string extension = Path.GetExtension(filename);
                if (extension == ".xls" || extension == ".xlsx")
                {
                    ExcelReader db = new ExcelReader(filename, true, false);
                    TableSelectDialog t = new TableSelectDialog(db.GetWorksheetList());

                    if (t.ShowDialog(this) == DialogResult.OK)
                    {
                        DataTable tableSource = db.GetWorksheet(t.Selection);
                        this.dgvTestingSource.DataSource = tableSource;
                        myTestDT = tableSource.Copy();
                    }
                }
            }
        }
    }
}

程序可以参考：https://github.com/accord-net/framework/blob/development/Unit%20Tests/Accord.Tests.MachineLearning/DecisionTrees/C45LearningTest.cs