/**
* get the vector of one instance
*
* @param p_Instance
* @param p_Stat
* @return
*/
private svm_node[] getVector(IInstance p_Instance, IStatistic p_Stat, int[][] p_Indice) {
String value = null;
int kIndex = 0;
int featureSize = p_Instance.size();
Hashtable<Integer, Double> exist = new Hashtable<Integer, Double>();
for (int fIndex = 0; fIndex < featureSize; fIndex++) {
IFeature feature = p_Instance.getFeature(fIndex);
kIndex = p_Stat.getIndex(feature.getKey());
if (kIndex < 0) {
continue;
}
if (ANumericFeature.class.isInstance(feature)) {
exist.put(p_Indice[kIndex][0], Double.parseDouble(feature.getValue()));
} else if (ABinaryFeature.class.isInstance(feature)) {
if (feature.getValue().equals("1")) {
exist.put(p_Indice[kIndex][0], 1.0);
}
} else {
List<String> values = p_Stat.getValue(kIndex);
value = feature.getValue();
if (value == null || !p_Stat.contains(kIndex, value)) {
value = p_Stat.getDefaultValue();
}
for (int i = 0; i < values.size(); i++) {
if (values.get(i).equals(value)) {
exist.put(p_Indice[kIndex][i], 1.0);
break;
}
}
}
}
ArrayList<Integer> indice = new ArrayList<Integer>(exist.keySet());
Collections.sort(indice);
svm_node[] retVal = new svm_node[exist.size()];
for (int i = 0; i < indice.size(); i++) {
retVal[i] = new svm_node();
retVal[i].index = indice.get(i);
retVal[i].value = exist.get(indice.get(i));
}
return retVal;
}
/**
* load the statistic of p_iLexelt for each feature type in statistic if feature is binary keep it
* else if feature is list and the number of values is less than 2 one new feature else set each
* value as a new feature
*
* @param p_iLexelt lexelt
* @return indices
* @throws ClassNotFoundException cannot find the defined type
*/
protected int[][] loadStatistic(ILexelt p_iLexelt) throws ClassNotFoundException {
int[][] retIndice = null;
int accuIndex = 1;
if (p_iLexelt != null) {
IStatistic stat = p_iLexelt.getStatistic();
int keySize = stat.getKeys().size();
retIndice = new int[keySize][0];
int keyIndex = 0;
for (keyIndex = 0; keyIndex < keySize; keyIndex++) {
Class<?> type = Class.forName(stat.getType(keyIndex));
if (ANumericFeature.class.isAssignableFrom(type)
|| ABinaryFeature.class.isAssignableFrom(type)) {
retIndice[keyIndex] = new int[] {accuIndex++};
} else {
List<String> values = stat.getValue(keyIndex);
retIndice[keyIndex] = new int[values.size()];
for (int i = 0; i < values.size(); i++) {
retIndice[keyIndex][i] = accuIndex++;
}
}
}
}
return retIndice;
}
/**
* change tags to integer (start from 1)
*
* @param p_Stat statistic
* @param p_Tags real tags
* @return new tags
*/
protected HashSet<Integer> processTags(IStatistic p_Stat, ArrayList<String> p_Tags) {
HashSet<Integer> retVal = new HashSet<Integer>();
if (p_Tags == null || p_Tags.size() == 0) {
retVal.add(0);
} else {
for (String tag : p_Tags) {
Integer iTag = 0;
if (!tag.equals("'?'") && !tag.equals("?")) {
iTag = p_Stat.getTagsInOrder().indexOf(tag);
if (iTag < 0) {
iTag = -1;
}
iTag++;
}
retVal.add(iTag);
}
}
return retVal;
}