我正在尝试学习如何在C＃中实现决策树。我发现的例子对我没有帮助。我要做的是为用户报告的问题提供解决方案，我有一个包含三千个问题和答案的数据库。我正在使用dotnet并尝试使用Accord.net框架，但是我没有得到它，我需要一个示例或有关实现方法的提示。
正如我所说的，我有一个包含旧问题和答案的数据库，该数据库将用于训练算法并提供对未来问题的可能答案。
附言我正在使用Visual Studio Community 2017。
最佳答案
帮助理解如何将这个问题映射到决策树问题的第一件事是：每个问题都应被视为决策树中的拆分节点（或决策节点），并为每个问题提供答案的集合，帮助形成每个拆分节点的子节点。在您的情况下，我们会将每个问题视为数据表的一列，并将表的每一行视为为每个问题提供的一组答案。
注意的第二件事是决策树需要预测“某物”，在您的情况下，它是对其中一个问题的答案，我们将其视为表中的最后一列。
由于没有提供书面示例，因此我假设是一个医疗数据库，其中包含一组问题和一组答案，如下面的代码所示。您可以轻松地将此映射到任何其他数据类型，只要它也是一组问题和答案即可。在下面给出的示例中，预测列/问题是已诊断的疾病。
[-此代码的灵感来自@ http://accord-framework.net/docs/html/T_Accord_MachineLearning_DecisionTrees_Learning_ID3Learning.htm]提供的示例
[-前提：您需要安装Accord.Net软件包：Install-Package Accord.MachineLearning]
using Accord.MachineLearning.DecisionTrees;
using Accord.MachineLearning.DecisionTrees.Learning;
using Accord.Math.Optimization.Losses;
using System;
using System.Data;
using Accord.Statistics.Filters;
using System.Linq;
using Accord.Math;
namespace DecisionTreeExamples
{
public class QandADecisionTree{
public static void QandADecisionTreeExample()
{
// Step1: Since there was no sample provided in this question, we first construct our data
DataTable data = new DataTable("QandA");
// Step2: Adding questions. Each question is a column in the data table.
PopulateQuestions(data); /*Note that the last column/question is the prediction column. */
// Step3: Adding the history of 6 patients and the corresponding diagnosis (This is also our training data)
PopulateAnswers(data);
//Step4: Next we need to convert our data table to its numerical representation, which is what ID3 algorithm expects as input.
Codification codification = new Codification(data);
DataTable codifiedData = codification.Apply(data);
//Note: specify all the columns except the last column, which is the value that the decision tree should predict,
int[][] input = codifiedData.ToJagged<int>("Are you above 50 years old?",
"What is your gender?",
"Do you have headache?",
"Has your blood pressure gone above 12 in the last 30 days?",
"Do you have a running nose?",
"Do you have a soat throat?",
"Do you have fever?");
int[] predictions = codifiedData.ToArray<int>("Diagnosed Disease");
// Step4: produce Decision tree.
ID3Learning decisionTreeLearningAlgorithm = new ID3Learning { };
DecisionTree decisionTree = decisionTreeLearningAlgorithm.Learn(input, predictions);
// Step5: Use the produced decision tree on a test data
int[] query = codification.Transform(new[,]{ {"Are you above 50 years old?","no" },
{ "What is your gender?","male" },
{ "Do you have headache?" ,"yes"},
{ "Has your blood pressure gone above 12 in the last 30 days?","no" },
{ "Do you have a running nose?","yes" },
{ "Do you have a soat throat?" ,"yes"},
{ "Do you have fever?" ,"yes"} });
int result = decisionTree.Decide(query);
string diagnosis = codification.Revert("Diagnosed Disease", result);
Console.WriteLine($"Diagnosed disease: {diagnosis}"); // Prints Common Cold
}
/// <summary>
/// Populates data table with questions, that in this example each question happens to be a column.
/// </summary>
/// <param name="data">Specifies the data base in this example which is a collection of questions and answers.</param>
private static void PopulateQuestions(DataTable data)
{
data.Columns.Add(new DataColumn("Are you above 50 years old?", typeof(string))); //Q1
data.Columns.Add(new DataColumn("What is your gender?", typeof(string))); //Q2
data.Columns.Add(new DataColumn("Do you have headache?", typeof(string))); //Q3
data.Columns.Add(new DataColumn("Has your blood pressure gone above 12 in the last 30 days?", typeof(string))); //Q4
data.Columns.Add(new DataColumn("Do you have a running nose?", typeof(string))); //Q5
data.Columns.Add(new DataColumn("Do you have a soat throat?", typeof(string))); //Q6
data.Columns.Add(new DataColumn("Do you have fever?", typeof(string))); //Q7
data.Columns.Add(new DataColumn("Diagnosed Disease", typeof(string))); // Prediction Column
}
/// <summary>
/// Populates data table with rows, that in this example each row happens to represent a patient, and each column, except the last one is a question. The value of a column in a row means the reponse patient provided to that question.
/// </summary>
/// <param name="data">Specifies the data base in this example which is a collection of questions and answers.</param>
private static void PopulateAnswers(DataTable data)
{
var patient1 = data.NewRow();
patient1["Are you above 50 years old?"] = "no";
patient1["What is your gender?"] = "male";
patient1["Do you have headache?"] = "yes";
patient1["Has your blood pressure gone above 12 in the last 30 days?"] = "no";
patient1["Do you have a running nose?"] = "yes";
patient1["Do you have a soat throat?"] = "yes";
patient1["Do you have fever?"] = "yes";
patient1["Diagnosed Disease"] = "Common Cold";
data.Rows.Add(patient1);
var patient2 = data.NewRow();
patient2["Are you above 50 years old?"] = "yes";
patient2["What is your gender?"] = "female";
patient2["Do you have headache?"] = "yes";
patient2["Has your blood pressure gone above 12 in the last 30 days?"] = "yes";
patient2["Do you have a running nose?"] = "no";
patient2["Do you have a soat throat?"] = "no";
patient2["Do you have fever?"] = "no";
patient2["Diagnosed Disease"] = "High Blood Pressure";
data.Rows.Add(patient2);
var patient3 = data.NewRow();
patient3["Are you above 50 years old?"] = "yes";
patient3["What is your gender?"] = "male";
patient3["Do you have headache?"] = "yes";
patient3["Has your blood pressure gone above 12 in the last 30 days?"] = "yes";
patient3["Do you have a running nose?"] = "no";
patient3["Do you have a soat throat?"] = "no";
patient3["Do you have fever?"] = "no";
patient3["Diagnosed Disease"] = "High Blood Pressure";
data.Rows.Add(patient3);
var patient4 = data.NewRow();
patient4["Are you above 50 years old?"] = "no";
patient4["What is your gender?"] = "female";
patient4["Do you have headache?"] = "yes";
patient4["Has your blood pressure gone above 12 in the last 30 days?"] = "no";
patient4["Do you have a running nose?"] = "yes";
patient4["Do you have a soat throat?"] = "yes";
patient4["Do you have fever?"] = "yes";
patient4["Diagnosed Disease"] = "Common Cold";
data.Rows.Add(patient4);
var patient5 = data.NewRow();
patient5["Are you above 50 years old?"] = "yes";
patient5["What is your gender?"] = "female";
patient5["Do you have headache?"] = "yes";
patient5["Has your blood pressure gone above 12 in the last 30 days?"] = "no";
patient5["Do you have a running nose?"] = "yes";
patient5["Do you have a soat throat?"] = "yes";
patient5["Do you have fever?"] = "yes";
patient5["Diagnosed Disease"] = "Common Cold";
data.Rows.Add(patient5);
var patient6 = data.NewRow();
patient6["Are you above 50 years old?"] = "yes";
patient6["What is your gender?"] = "female";
patient6["Do you have headache?"] = "no";
patient6["Has your blood pressure gone above 12 in the last 30 days?"] = "yes";
patient6["Do you have a running nose?"] = "no";
patient6["Do you have a soat throat?"] = "no";
patient6["Do you have fever?"] = "no";
patient6["Diagnosed Disease"] = "High Blood Pressure";
data.Rows.Add(patient6);
}
}
}
关于c# - 使用Accord.Net框架的决策树，我们在Stack Overflow上找到一个类似的问题：https://stackoverflow.com/questions/47082295/