public static Node1 FindBestSplit(Node1 node, List <int> nums, int[] isUsed)
{
// 计算node子树系统的信息熵
double entropy = CalEntropy(node.classCount, node.count);
if (IfStop(node, entropy, isUsed))
{
return(node); // 停止分裂，对node做适当的赋值以标记分裂停止，返回这个节点
}
var info = new SplitInfo1();
var countAll = node.count;
for (int i = 0; i < isUsed.Length - 1; i++)
{
if (isUsed[i] == 1)
{
continue;
}
if (types[i] == 0) // 离散型属性值
{
}
}
return(null);
}
/// <summary>
/// 是否停止分裂
/// </summary>
/// <param name="node"></param>
/// <param name="entropy"></param>
/// <param name="isUsed"></param>
/// <returns></returns>
public static bool IfStop(Node1 node, double entropy, int[] isUsed)
{
var counts = node.classCount;
var countAll = node.count;
int maxIndex = 0; // 占比最大的分类值在 classCount 数组中的索引
int deep = node.deep;
// 最后一项表示分类，而非输入特征属性
bool flag = true; // 如果输入属性已经用完
for (int i = 0; i < isUsed.Length - 1; i++)
{
if (isUsed[i] == 0)
{
flag = false; // 输入属性尚未用完
break;
}
}
if (deep >= maxDeep || entropy == 0 || countAll < LimitCount || flag)
{
maxIndex = node.result + 1;
node.feature_type = "result";
node.features = new List <string>()
{
maxIndex + ""
};
node.leafError = countAll - (int)counts[node.result];
node.leafCount = 1;
return(true);
}
return(false);
}
