/*********************************************************************** This file is part of KEEL-software, the Data Mining tool for regression, classification, clustering, pattern mining and so on. Copyright (C) 2004-2010 F. Herrera (herrera@decsai.ugr.es) L. S�nchez (luciano@uniovi.es) J. Alcal�-Fdez (jalcala@decsai.ugr.es) S. Garc�a (sglopez@ujaen.es) A. Fern�ndez (alberto.fernandez@ujaen.es) J. Luengo (julianlm@decsai.ugr.es) This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see http://www.gnu.org/licenses/ **********************************************************************/ /** * <p> * @author Written by Antonio Alejandro Tortosa (University of Granada) 01/07/2008 * @author Modified by Xavi Sol� (La Salle, Ram�n Llull University - Barcelona) 16/12/2008 * @version 1.1 * @since JDK1.2 * </p> */ package keel.Algorithms.Rule_Learning.C45RulesSA; import java.util.Vector; public class Rule { /** * <p> * Representation of a string of simple rules chained by 'and's: <b>exemple[a1][=|>|<=]v1 && exemple[a2][=|>=|<=]v2</b> * </p> */ // operator > public static int GREATER=SimpleRule.GREATER; //operator <= public static int LOWER=SimpleRule.LOWER; //operator = public static int EQUAL=SimpleRule.EQUAL; //string of simple rules private Vector chain; //right side of the rule private String type; /** The redundancy factor in theory description length */ private static double REDUNDANCY_FACTOR = 0.5; /** The theory weight in the MDL calculation */ private double MDL_THEORY_WEIGHT = 1.0; /***************Private methods**********************/ /** * It returns wether a simple rule is part of the rule * @param sr SimpleRule the simple rule * @return true if the simple rule is part of the rule */ private boolean contains(SimpleRule sr){ boolean is_contained=false; for (int i=0;i<chain.size()&&!is_contained;i++) is_contained=sr.isEqual((SimpleRule) chain.elementAt(i)); return is_contained; } /***************Private methods**********************/ /** * Constructs an empty rule. */ public Rule() { chain=new Vector(); type=""; } /** * Returns the number of the instances covered by the rule in a given dataset. * This method allows to ignore a simple rule from that rule. * @param data MyDataset the dataset * @param select Mask the mask with the active entries of the dataset * @param ignore int id of the single rule that it will be ignore in the applying of the rule * @return number of instances (from the active entries) covered by the rule */ public int apply(MyDataset data,Mask select,int ignore){ if (this.chain.size()==0) return select.getnActive(); if (this.chain.size()==1 && ignore==0) return select.getnActive(); int output=0; //That variable will take the count of the covered entries select.resetIndex(); while (select.next()){ double[] exemple=data.getExample(select); boolean salir=false; for (int j=0;j<chain.size() && !salir;j++){ SimpleRule sr = (SimpleRule) chain.elementAt(j); int attribute = sr.getAttribute(); double value = sr.getValue(); if (data.isMissing(select, attribute) && j != ignore) { salir = true; //if any value is missing the whole comprobation for that entry fails } else if (j != ignore) { if (sr.isDiscret()) { salir = ! ( exemple[attribute] == value); } else { if (sr.getOperator() == SimpleRule.GREATER) salir = ! ( exemple[attribute] > value); else salir = ! ( exemple[attribute] <= value); } //end if (sr.isDiscret()) } //end if (j!ignore) }//end for if (!salir) output++; }//end while return output; } /** * Returns the number of the instances covered by the rule in a given dataset * @param data MyDataset the dataset * @param select Mask the mask with the active entries of the dataset * @return number of instances (from the active entries) covered by the rule */ public int apply (MyDataset data,Mask select){ return apply(data,select,-1); } /** * It returns the number of the instances covered by the rule in a given dataset * @param data MyDataset the dataset * @return number of instances (from the active entries) covered by the rule */ public int apply (MyDataset data){ return apply(data,new Mask(data.size()),-1); } /** * It returns the number of true positives,true negatives,false positives and false negatives of the rule in a given dataset * @param data MyDataset the dataset * @param positives active positive instances of data * @param negatives active negative instances of data * @return number of true positives, false positives, true negatives and false negatives of the rule in the following order: {tp,tn,fp,fn} */ public Stats apply (MyDataset data,Mask positives,Mask negatives){ Stats stats=new Stats(); stats.tp=apply(data,positives); //true positives stats.fn=positives.getnActive()-stats.tp; //false negatives stats.fp=apply(data,negatives); //false positives stats.tn=negatives.getnActive()-stats.fp; //true negatives return stats; } /** * Returns the Minimum Data Length of a dataset given a theory (this rule). See [Quinlan95] * @param data MyDataset the datasets * @param positives Mask active positive entries of data * @param negatives Mask active negative entries of data * @return the MDL of data given this rule */ public double getExceptionCost(MyDataset data,Mask positives,Mask negatives){ Stats quartet=apply(data,positives,negatives); double tp=quartet.tp,tn=quartet.tn,fp=quartet.fp,fn=quartet.fn; double tp_prob,tn_prob,fp_prob,fn_prob; double U=tn+fn,C=tp+fp; //uncovered & covered cases double D=U+C,e=fn+fp; if ( C > 0.5 * (D) ) { return Utilities.log2(D+1) + biased(C, fp, 0.5 * e) + biased(U, fn, fn); } else { return Utilities.log2(D+1) + biased(C, fp, fp) + biased(U, fn, 0.5 * e); } } /** * Static version. * Returns the Minimum Data Length of a dataset given a theory (this rule). See [Quinlan95] * @param data MyDataset the datasets * @param tp int true positives * @param tn int true negatives * @param fp int false positives * @param fn int false negatives * @return the MDL of data given this rule */ public static double getExceptionCost(MyDataset data,int tp,int tn,int fp,int fn){ double tp_prob,tn_prob,fp_prob,fn_prob; double U=tn+fn,C=tp+fp; //uncovered & covered cases double D=U+C,e=fn+fp; if (C==0) return Double.MAX_VALUE; if ( C > 0.5 * (D) ) { return Utilities.log2(D+1) + biased(C, fp, 0.5 * e) + biased(U, fn, fn); } else { return Utilities.log2(D+1) + biased(C, fp, fp) + biased(U, fn, 0.5 * e); } } public static double biased(double N, double E, double ExpE){ double Rate; if ( ExpE <= 1E-6 ) { return ( E == 0 ? 0.0 : 1E6 ); } else if ( ExpE >= N-1E-6 ) { return ( E == N ? 0.0 : 1E6 ); } Rate = ExpE/N; return -E * Utilities.log2(Rate) - (N-E) * Utilities.log2(1-Rate); } /** * Subset description length: <br> * S(t,k,p) = -k*log2(p)-(n-k)log2(1-p) * * Details see Quilan: "MDL and categorical theories (Continued)",ML95 * * @param t the number of elements in a known set * @param k the number of elements in a subset * @param p the expected proportion of subset known by recipient * @return the subset description length */ public static double subsetDL(double t, double k, double p){ double rt = (p>0.0) ? (- k*Utilities.log2(p)) : 0.0; rt -= (t-k)*Utilities.log2(1-p); return rt; } /** * The description length of the theory for a given rule. Computed as:<br> * 0.5* [||k||+ S(t, k, k/t)]<br> * where k is the number of antecedents of the rule; t is the total * possible antecedents that could appear in a rule; ||K|| is the * universal prior for k , log2*(k) and S(t,k,p) = -k*log2(p)-(n-k)log2(1-p) * is the subset encoding length.<p> * * Details see Quilan: "MDL and categorical theories (Continued)",ML95 * * @param data MyDataset the dataset * @return the theory DL, weighted if weight != 1.0 */ public double theoryDL(MyDataset data){ double k = size(); if(k == 0) return 0.0; double tdl = Math.log(k); if(k > 1) // Approximation tdl += 2.0 * Math.log(tdl); // of log2 star double totalCond=0.0; for (int i=0;i<k;i++) totalCond+=data.numAllConditions(getSimpleRule(i).getAttribute()); tdl += subsetDL(totalCond, k, k/totalCond); return MDL_THEORY_WEIGHT * REDUNDANCY_FACTOR * tdl; } /** * Returns the i-ieth simple rule of this rule. * @param i position of the simple rule * @return the i-ieth simple rule of this rule. */ public SimpleRule getSimpleRule(int i){ return (SimpleRule) chain.elementAt(i); } /** * Adds a simple rule to this rule. * @param attribute int attribute id (position of the attribute) * @param value double attribute's value * @param operator int rule operator */ public void grow(int attribute,double value,int operator){ SimpleRule sr=new SimpleRule(attribute,value,operator); chain.add(sr); } /** * Adds a simple rule to this rule. * @param sr SimpleRule the simple rule */ public void grow(SimpleRule sr){ chain.add(sr); } /** * It sets the right side of the rule. * @param new_class double new class of the rule */ public void setType(String new_class){ this.type=new_class; } /** * It returns the right side (class) of the rule. * @return the right side (class) of the rule. */ public String getType(){ return type; } /** * It returns a copy of this rule * @return a copy of this rule */ public Rule getCopy(){ Rule r=new Rule(); for (int i=0;i<chain.size();i++) r.grow(this.getSimpleRule(i).getCopy()); return r; } /** * Deletes a simple rule from this chain * @param pos int position of the simple rule of the rule */ public void prune(int pos){ chain.remove(pos); } /** * Returns the size (number of simple rules) of the rule * @return the size (number of simple rules) of the rule */ public int size(){ return chain.size(); } /** * Return wether this rule is equal to another given rule * @param r Rule the given rule * @return true if this rule is equal to the given rule */ public boolean isEqual(Rule r){ if (chain.size()!=r.size()) return false; boolean is_equal=true; for (int i=0;i<r.size() && is_equal;i++) is_equal=this.contains((SimpleRule) r.getSimpleRule(i)); return is_equal; } /** * Returns a string representation of this Rule, containing the String representation of each SimpleRule. * @return a string representation of this Rule, containing the String representation of each SimpleRule. */ public String toString(){ String output="("; if (chain.size()!=0){ output+=((SimpleRule)chain.elementAt(0)).toString(); } for (int i=1;i<chain.size();i++) output+=" && "+((SimpleRule)chain.elementAt(i)).toString(); output+=")"; if (!type.equals("")){ output+="-> "; output+=type; } return output; } }