// regular DM method
private static void regular() {
List<Point> positiveTrainData = new ArrayList<Point>();
// select the minority class instances
for (Point point : dataSet) {
if (point.getLabel() == 0) positiveTrainData.add(point);
}
System.out.print("train data :" + dataSet.size() + "\t");
System.out.println("train positive :" + positiveTrainData.size());
// 将训练集进行smote操作
smote(positiveTrainData);
// generate new dataset
String trainFileName = NAME + "SMOTETrain" + ".arff";
String testFileName = NAME + "SMOTETest" + ".arff";
Generate.generate(dataSet, pointSet, COL, fileName, trainFileName);
Generate.generate(testSet, new ArrayList<Point>(), COL, fileName, testFileName);
pointSet.clear();
classify(trainFileName, testFileName);
// 不进行任何处理
trainFileName = NAME + "TrainWS" + ".arff";
testFileName = NAME + "TestWS" + ".arff";
Generate.generate(dataSet, new ArrayList<Point>(), COL, fileName, trainFileName);
Generate.generate(testSet, new ArrayList<Point>(), COL, fileName, testFileName);
classify(trainFileName, testFileName);
}
/** divide the dataset into k folds */
public static void crossValidation() {
// double sum = 0;
// ---------------------分为k折-----------------------------
// 初始化为k fold
for (int i = 0; i < FOLD; i++) {
ArrayList<Point> tmp = new ArrayList<Point>();
allData.add(tmp);
}
// 选一个 删一个
List<Integer> chosen = new ArrayList<Integer>();
for (int i = 0; i < dataSet.size(); i++) {
chosen.add(i);
}
for (int i = 0; i < FOLD; i++) {
int choose = 0;
while (choose < ROW / FOLD && i != FOLD - 1) {
int rand = new Random().nextInt(dataSet.size());
if (chosen.contains(rand)) {
chosen.remove(new Integer(rand));
choose++;
allData.get(i).add(dataSet.get(rand));
}
}
// 最后一折全部添加
if (i == FOLD - 1) {
for (Integer o : chosen) {
allData.get(i).add(dataSet.get(o));
}
}
}
// ------------------取一折为测试，其余为训练集-----------------------------
for (int fold = 0; fold < FOLD; fold++) {
List<Point> trainData = new ArrayList<Point>();
List<Point> testData = new ArrayList<Point>();
List<Point> positiveTrainData = new ArrayList<Point>();
List<Point> positiveTestData = new ArrayList<Point>();
testData.addAll(allData.get(fold));
for (List<Point> ps : allData) {
if (ps != allData.get(fold)) trainData.addAll(ps);
}
// select the minority class instances
for (Point point : trainData) {
if (point.getLabel() == 0) positiveTrainData.add(point);
}
System.out.print("train data :" + trainData.size() + "\t");
System.out.println("train positive :" + positiveTrainData.size());
for (Point point : testData) {
if (point.getLabel() == 0) positiveTestData.add(point);
}
System.out.print("test data :" + testData.size() + "\t");
System.out.println("test positive :" + positiveTestData.size());
// 不进行任何处理
// System.out.println("----------------Original result-----------------------");
String trainFileName = NAME + "OriginalTrain" + fold + ".arff";
String testFileName = NAME + "OriginalTest" + fold + ".arff";
// Generate.generate(trainData,new ArrayList<Point>(),COL,fileName,trainFileName);
// Generate.generate(testData,new ArrayList<Point>(),COL,fileName,testFileName);
//
// classify(trainFileName,testFileName);
// generate new dataset
// 将训练集进行smote操作
smote(positiveTrainData);
System.out.println("----------------SMOTE result-----------------------");
trainFileName = NAME + "SMOTETrain" + fold + ".arff";
testFileName = NAME + "SMOTETest" + fold + ".arff";
Generate.generate(trainData, pointSet, COL, fileName, trainFileName);
Generate.generate(testData, new ArrayList<Point>(), COL, fileName, testFileName);
pointSet.clear();
classify(trainFileName, testFileName);
System.out.println("-----------------------------------------------");
}
// System.out.println("average precision is :" + sum/FOLD);
}
