parameter.java

来自「Fast implementation of C4/5 in Java」· Java 代码 · 共 49 行

/**
 * @(#)Parameter.java        1.5.0 09/01/18
 */

package ml.classifier.dt;

/**
 * The default parameter setting in the implementation of C4.5 algorithm
 *
 * @author Ping He
 * @author Xiaohua Xu
 */

public class Parameter {
	/**
	 * Its minus is the default minimal valid Gain value.
	 */
	public static final float EPSILON = 0.001f;
	/**
	 * The default precision in comparing two float values.
	 */
	public static final double PRECISION = 1E-5;
	/**
	 * The default minimal weight to construct an internal tree node.
	 */
	public static final float MINWEIGHT = 2.0f;

	/**
	 * The default confidence level in c4.5's error-based pruning. 
	 * <p>
	 * CONFIDENCE must be changed together with SQUAREDSTD parameter.
	 * The setConfidence(float confidence) method in ml.classifier.dt.Estimator  
	 * can be used to reset confidence and its corresponding SQUAREDSTD.
	 * </p>
	 */
	public static final float CONFIDENCE = 0.25f;
	/**
	 * The default squared standard deviation value corresponding to the default CONFIDENCE.
	 * <p>
	 * SQUAREDSTD must be changed together with CONFIDENCE parameter.
	 * The setConfidence(float confidence) method in ml.classifier.dt.Estimator  
	 * can be used to reset confidence and its corresponding SQUAREDSTD.
	 * </p>
	 */
	public static final float SQUAREDSTD = 0.47955623f;
	
	private Parameter(){
	}
}