Ai算法java代码示例

以下是一个简单的基于Java的决策树算法示例：

import java.util.ArrayList;

public class DecisionTree {
  private TreeNode root;
  
  public DecisionTree() {
    root = null;
  }
  
  public void train(ArrayList<DataPoint> data) {
    root = buildTree(data);
  }
  
  public String predict(DataPoint data) {
    return predictHelper(root, data);
  }
  
  private TreeNode buildTree(ArrayList<DataPoint> data) {
    if (data.isEmpty()) {
      return new TreeNode("unknown");
    }
    
    // 选择最优的特征作为分裂点
    Feature bestFeature = getBestFeature(data);
    TreeNode node = new TreeNode(bestFeature.getName());
    
    // 递归构建子树
    ArrayList<String> featureValues = bestFeature.getValues();
    for (String value : featureValues) {
      ArrayList<DataPoint> subset = getSubset(data, bestFeature, value);
      TreeNode child = buildTree(subset);
      node.addChild(value, child);
    }
    
    return node;
  }
  
  private String predictHelper(TreeNode node, DataPoint data) {
    if (node.isLeaf()) {
      return node.getLabel();
    }
    
    String featureValue = data.getFeature(node.getName());
    TreeNode child = node.getChild(featureValue);
    return predictHelper(child, data);
  }
  
  private Feature getBestFeature(ArrayList<DataPoint> data) {
    // 计算信息增益，选择信息增益最大的特征
    double entropy = getEntropy(data);
    Feature bestFeature = null;
    double bestGain = 0;
    for (Feature feature : data.get(0).getFeatures()) {
      double gain = entropy - getConditionalEntropy(data, feature);
      if (gain > bestGain) {
        bestGain = gain;
        bestFeature = feature;
      }
    }
    
    return bestFeature;
  }
  
  private double getEntropy(ArrayList<DataPoint> data) {
    // 计算数据集的熵
    int size = data.size();
    int[] counts = new int[2];
    for (DataPoint point : data) {
      counts[point.getLabel()]++;
    }
    
    double entropy = 0;
    for (int count : counts) {
      if (count > 0) {
        double p = (double) count / size;
        entropy -= p * Math.log(p) / Math.log(2);
      }
    }
    
    return entropy;
  }
  
  private double getConditionalEntropy(ArrayList<DataPoint> data, Feature feature) {
    // 计算特征条件下的数据集的熵
    double conditionalEntropy = 0;
    ArrayList<String> featureValues = feature.getValues();
    for (String value : featureValues) {
      ArrayList<DataPoint> subset = getSubset(data, feature, value);
      double p = (double) subset.size() / data.size();
      conditionalEntropy += p * getEntropy(subset);
    }
    
    return conditionalEntropy;
  }
  
  private ArrayList<DataPoint> getSubset(ArrayList<DataPoint> data, Feature feature, String value) {
    // 获取某个特征取某个值的数据子集
    ArrayList<DataPoint> subset = new ArrayList<DataPoint>();
    for (DataPoint point : data) {
      if (point.getFeature(feature.getName()).equals(value)) {
        subset.add(point);
      }
    }
    
    return subset;
  }
}

class TreeNode {
  private String name;
  private String label;
  private ArrayList<TreeNode> children;
  private ArrayList<String> values;
  
  public TreeNode(String name) {
    this.name = name;
    label = null;
    children = new ArrayList<TreeNode>();
    values = new ArrayList<String>();
  }
  
  public void addChild(String value, TreeNode child) {
    children.add(child);
    values.add(value);
  }
  
  public String getName() {
    return name;
  }
  
  public boolean isLeaf() {
    return label != null;
  }
  
  public void setLabel(String label) {
    this.label = label;
  }
  
  public String getLabel() {
    return label;
  }
  
  public TreeNode getChild(String value) {
    int index = values.indexOf(value);
    return children.get(index);
  }
}

class DataPoint {
  private ArrayList<Feature> features;
  private int label;
  
  public DataPoint(ArrayList<Feature> features, int label) {
    this.features = features;
    this.label = label;
  }
  
  public String getFeature(String name) {
    for (Feature feature : features) {
      if (feature.getName().equals(name)) {
        return feature.getValue();
      }
    }
    
    return null;
  }
  
  public ArrayList<Feature> getFeatures() {
    return features;
  }
  
  public int getLabel() {
    return label;
  }
}

class Feature {
  private String name;
  private String value;
  private ArrayList<String> values;
  
  public Feature(String name, String value) {
    this.name = name;
    this.value = value;
    values = new ArrayList<String>();
    values.add(value);
  }
  
  public Feature(String name, ArrayList<String> values) {
    this.name = name;
    this.values = values;
    value = null;
  }
  
  public String getName() {
    return name;
  }
  
  public String getValue() {
    return value;
  }
  
  public ArrayList<String> getValues() {
    return values;
  }
}

这段代码实现了一个简单的决策树算法，包括训练和预测功能。它使用了三个类：`DecisionTree`表示决策树，`TreeNode`表示树节点，`DataPoint`表示数据点。其中，`DecisionTree`类中的`train`方法用于训练决策树，`predict`方法用于预测新的数据点。这段代码只是一个简单的示例，实际应用中可能需要进行更多的优化和改进。