c# - 使用ID3算法、Accord.Net框架进行预测

Question

我使用以下代码来查找给定商品在更改商品的价格、折扣和广告时的预期销售额。这是使用 Accord.Net 库使用 ID3 算法实现的。

namespace PnredictionSales
{
public partial class WebForm1 : System.Web.UI.Page
{
    protected void Page_Load(object sender, EventArgs e)
    {
        DataTable data = new DataTable("Sales prediction Example");

        data.Columns.Add("RowKey");
        data.Columns.Add("Brand");
        data.Columns.Add("PriceRange");
        data.Columns.Add("Discount");
        data.Columns.Add("Advertisement");
        data.Columns.Add("ExpSales");

        //  data.Columns.Add("Wind");
        //  data.Columns.Add("PlayTennis");

        data.Rows.Add("D1", "Highland", "R1", "yes", "No", "B");
        data.Rows.Add("D2", "Highland", "R1", "yes", "yes", "C");
        data.Rows.Add("D3", "Anchor", "R1", "yes", "No", "B");
        data.Rows.Add("D4", "Flora", "R2", "yes", "No", "B");
        data.Rows.Add("D5", "Flora", "R3", "No", "No", "A");
        data.Rows.Add("D6", "Flora", "R3", "No", "yes", "A");
        data.Rows.Add("D7", "Anchor", "R3", "No", "yes", "A");
        data.Rows.Add("D8", "Highland", "R2", "yes", "No", "B");
        data.Rows.Add("D9", "Highland", "R3", "No", "No", "A");
        data.Rows.Add("D10", "Flora", "R2", "No", "No", "B");
        data.Rows.Add("D11", "Highland", "R2", "No", "yes", "B");
        data.Rows.Add("D12", "Anchor", "R2", "yes", "yes", "A");
        data.Rows.Add("D13", "Anchor", "R1", "No", "No", "B");
        data.Rows.Add("D14", "Flora", "R2", "yes", "yes", "A");

        Codification codebook = new Codification(data);

        DecisionVariable[] attributes =
        {
            new DecisionVariable("Brand", 3),  new DecisionVariable("PriceRange",3),
            new DecisionVariable("Discount",2),new DecisionVariable("Advertisement",2) 
        };

        int classCount=3; // 2 possible output values for playing tennis: yes or no

        DecisionTree tree = new DecisionTree(attributes, classCount);

        // Create a new instance of the ID3 algorithm
        ID3Learning id3learning = new ID3Learning(tree);

        // Translate our training data into integer symbols using our codebook:
        DataTable symbols = codebook.Apply(data);
        int[][] inputs = symbols.ToIntArray("Brand", "PriceRange","Discount","Advertisement");
        int[] outputs = symbols.ToIntArray("ExpSales").GetColumn(0);

        // Learn the training instances!

        id3learning.Run(inputs, outputs);
        int[] query = codebook.Translate("Flora","R1","yes","No");

        int output = tree.Compute(query.ToDouble());

        string answer = codebook.Translate("ExpSales", output); // answer will be "No".
        Label1.Text = answer;
    }
}

我的问题是:

当我将任何字符串值放入 int[] query = codebook.Translate("fff","eee","ffg","qqq"); 时，它会给我一个输出。我想请问这是什么原因？我的做法有错吗？我还想知道在数据表中组织数据以获得准确结果的最低要求是什么。

Answer 1

我尝试运行你的代码 - 但我得到一个异常而不是任何输出。看着它，我认为问题在于，当您创建 Codification 时，您没有指定要包含哪些列，因此它包含 RowKey 列，然后没有任何内容排列。相反，使用以下命令创建编码:

Codification codebook = new Codification(data, "Brand", "PriceRange", "Discount", "Advertisement", "ExpSales");

然后它似乎就起作用了。

当我再次尝试您的 int[] query = codebook.Translate("fff","eee","ffg","qqq"); 示例时，我只是得到一个异常(因为这些值不存在) - 所以我认为你必须有一个异常处理程序来向你隐藏这些问题。

就获得准确结果的最少数据而言，这实际上取决于数据的复杂程度以及其中包含的噪声量。您需要针对一组数据训练模型，然后针对完全不同的数据集测试其准确性，以衡量其是否有效。