package org.fnlp.ml.classifier.bayes;
import gnu.trove.iterator.TIntFloatIterator;
import java.io.Serializable;
import org.fnlp.ml.types.Instance;
import org.fnlp.ml.types.InstanceSet;
import org.fnlp.ml.types.alphabet.AlphabetFactory;
import org.fnlp.ml.types.sv.HashSparseVector;
/**
*
* 特征频数统计结构
* 用于统计特征、类别、特征-类别的出现频数，可以用于词频统计
* @author sywu
*
*/
public class ItemFrequency implements Serializable{
private int size,featureSize,typeSize;
private int[] itemFrequency;
private int[] typeFrequency;
private int[] featureFrequency;
private int total=0;
public ItemFrequency(int featureSize,int typeSize){
init(featureSize, typeSize);
}
public ItemFrequency(InstanceSet instSet){
this(instSet,instSet.getAlphabetFactory());
}
public ItemFrequency(InstanceSet instSet,AlphabetFactory af){
this(af.getFeatureSize(),af.getLabelSize());
statistic(instSet);
}
public void init(int featureSize,int typeSize){
this.setFeatureSize(featureSize);
this.setTypeSize(typeSize);
size=featureSize*typeSize;
itemFrequency=new int[size];
typeFrequency=new int[typeSize];
featureFrequency=new int[featureSize];
total=0;
}
public int getTypeSize() {
return typeSize;
}
public void setTypeSize(int typeSize) {
this.typeSize = typeSize;
}
public int getFeatureSize() {
return featureSize;
}
public void setFeatureSize(int featureSize) {
this.featureSize = featureSize;
}
public int getFeatureFrequency(int index){
return featureFrequency[index];
}
public int getTypeFrequency(int index){
return typeFrequency[index];
}
public int getItemFrequency(int index){
return itemFrequency[index];
}
public int getItemFrequency(int feature,int type){
return itemFrequency[feature+type*featureSize];
}
public void setItemFrequency(int index,int frequency){
setItemFrequency(index%featureSize,index/featureSize,index,frequency);
}
public void setItemFrequency(int feature,int type,int frequency){
setItemFrequency(feature,type,feature+type*featureSize,frequency);
}
private void setItemFrequency(int feature,int type,int index,int frequency){//index=feature+type*featureSize;
int diff=frequency-itemFrequency[index];
featureFrequency[feature]+=diff;
typeFrequency[type]+=diff;
itemFrequency[index]=frequency;
total+=diff;
}
public void addItemFrequency(int index,int diff){
addItemFrequency(index%featureSize,index/featureSize,index,diff);
}
public void addItemFrequency(int feature,int type,int diff){
addItemFrequency(feature,type,feature+type*featureSize,diff);
}
private void addItemFrequency(int feature,int type,int index,int diff){//index=feature+type*featureSize;
featureFrequency[feature]+=diff;
typeFrequency[type]+=diff;
itemFrequency[index]+=diff;
total+=diff;
}
public int getTotal() {
return total;
}
public void setTotal(int total) {
this.total = total;
}
/**
* 统计数据集词频
* @param set 数据集
*/
public void statistic(InstanceSet instSet){
int numSamples = instSet.size();
for(int i=0;i<numSamples;i++){
if(i%1000==0)
System.out.println((i+0.0)/numSamples);
Instance inst=instSet.get(i);
oneStepStatistic(inst);
}
}
private boolean oneStepStatistic(Instance inst) {
if(inst==null)
return false;
int[] type;
Object t=inst.getTarget();
if(t instanceof Integer){
type=new int[1];
type[0]=Integer.parseInt(t.toString());
}
else{
return false;
}
HashSparseVector data = (HashSparseVector) inst.getData();
TIntFloatIterator it = data.data.iterator();
while (it.hasNext()) {
it.advance();
int feature=it.key();
for(int i=0;i<type.length;i++){
addItemFrequency(feature, type[i], (int)it.value());
}
}
return true;
}
}