public static void generateTrainingDataFromFile(
String
fileLocation) // Requires that the original file had the metadata and requires that this
// file is formated the same in first sheet
{
testDataLL = (LinkedList<String[]>) dataLL.clone();
actualClassifications = (LinkedList<String>) classificationsLL.clone();
FileInputStream file;
try {
file = new FileInputStream(new File(fileLocation));
Workbook excelFile = new HSSFWorkbook(file);
Sheet sheet1 = excelFile.getSheetAt(0); // Data sheet
for (Row row : sheet1) {
String data[] = new String[row.getPhysicalNumberOfCells() - 1];
String classification = "";
int offset =
0; // Used so that we can declare an array of the size of the attributes without the
// classification
for (Cell cell : row) {
int index = cell.getColumnIndex();
if (classificationLocation != index) {
data[index - offset] = cell.toString();
} else {
classification = cell.toString();
offset++;
}
}
// Even though data and classifications are not really used add it onto the end so it is
// still complete for in the event they end up being used in a later version
dataLL.add(data);
classificationsLL.add(classification);
trainingDataLL.add(data);
knownClassifications.add(classification);
// Check to see if we have seen that classification yet
int occurrences = 0;
for (int i = 0; i < classificationTypes.size() && occurrences == 0; i++) {
if (classificationTypes.get(i).compareTo(classification) == 0) {
occurrences = 1;
}
}
if (occurrences == 0) {
classificationTypes.add(classification);
}
}
excelFile.close();
} catch (FileNotFoundException e) {
System.out.println("Error file not found");
System.exit(0);
} catch (IOException e) {
System.out.println("Unable to read file, disk drive may be failing");
e.printStackTrace();
System.exit(0);
}
}
public static void writeExcelFile(String fileLocation) {
try {
FileOutputStream fileOut = new FileOutputStream(fileLocation);
HSSFWorkbook workbook = new HSSFWorkbook();
Font bold = workbook.createFont(); // Create font
bold.setBoldweight(Font.BOLDWEIGHT_BOLD); // Make font bold
CellStyle correctCell = workbook.createCellStyle();
correctCell.setFillForegroundColor(HSSFColor.GREEN.index);
correctCell.setFillBackgroundColor(HSSFColor.GREEN.index);
correctCell.setFillPattern(CellStyle.SOLID_FOREGROUND);
CellStyle incorrectCell = workbook.createCellStyle();
incorrectCell.setFillForegroundColor(HSSFColor.RED.index);
incorrectCell.setFillBackgroundColor(HSSFColor.RED.index);
incorrectCell.setFillPattern(CellStyle.SOLID_FOREGROUND);
CellStyle classificationCells = workbook.createCellStyle();
classificationCells.setFillForegroundColor(HSSFColor.YELLOW.index);
classificationCells.setFillBackgroundColor(HSSFColor.YELLOW.index);
classificationCells.setFillPattern(CellStyle.SOLID_FOREGROUND);
CellStyle attributeNameCells = workbook.createCellStyle();
attributeNameCells.setFont(bold);
CellStyle classificationAttributeCell = workbook.createCellStyle();
classificationAttributeCell.setFillForegroundColor(HSSFColor.YELLOW.index);
classificationAttributeCell.setFillBackgroundColor(HSSFColor.YELLOW.index);
classificationAttributeCell.setFillPattern(CellStyle.SOLID_FOREGROUND);
classificationAttributeCell.setFont(bold);
Sheet worksheet = workbook.createSheet("Results");
Row currRow = worksheet.createRow(0);
for (int attribute = 0; attribute < metadataLL.size() + 1; attribute++) {
Cell currCell = currRow.createCell(attribute);
if (attribute < metadataLL.size()) {
currCell.setCellValue(metadataLL.get(attribute)[0]);
if (metadataLL.get(attribute)[2].compareTo("classifier") == 0) {
currCell.setCellStyle(classificationAttributeCell);
} else {
currCell.setCellStyle(attributeNameCells);
}
} else {
currCell.setCellValue("Guessed Classification");
currCell.setCellStyle(attributeNameCells);
}
}
int correct = 0;
int incorrect = 0;
for (int node = 0; node < testDataLL.size(); node++) {
currRow = worksheet.createRow(node + 1); // Offset by one since first row is header data
int classifierCompleted = 0; // Necessary for if data does not end in classifier
for (int attribute = 0;
attribute < metadataLL.size() + 2;
attribute++) // +1 for the row for guessed data +1 for the row that contains
{
Cell currCell = currRow.createCell(attribute);
if (attribute < metadataLL.size()) // Print testingData
{
if (metadataLL.get(attribute)[2].compareTo("classifier") == 0) {
currCell.setCellValue(actualClassifications.get(node));
currCell.setCellStyle(classificationCells);
classifierCompleted++;
} else {
currCell.setCellValue(testDataLL.get(node)[attribute - classifierCompleted]);
}
} else if (attribute == metadataLL.size()) // Print guessed classification
{
currCell.setCellValue(guessedClassifications.get(node));
if (guessedClassifications.get(node).compareTo(actualClassifications.get(node)) == 0) {
currCell.setCellStyle(correctCell);
correct++;
} else {
currCell.setCellStyle(incorrectCell);
incorrect++;
}
if (node == testDataLL.size() - 1) // If this is the last loop
{
double precentRight = (double) correct / (correct + incorrect);
currRow = worksheet.createRow(node + 2);
currCell = currRow.createCell(attribute);
currCell.setCellValue(precentRight);
if (precentRight > .90) {
correctCell.setDataFormat(workbook.createDataFormat().getFormat("0.000%"));
currCell.setCellStyle(correctCell);
} else {
incorrectCell.setDataFormat(workbook.createDataFormat().getFormat("0.000%"));
currCell.setCellStyle(incorrectCell);
}
}
} else if (attribute
== metadataLL.size() + 1) // Print potential bad training data if the flag is true
{
if (unseenDataFlag.get(node)) {
currCell.setCellValue(
"This node an attribute value not in the training set, classifier selected is based on most frequent classifier. If laplacian smoothing is 1 or more this likely wont happen"); // TODO make this a bit shorter
}
}
}
}
worksheet = workbook.createSheet("Training Data");
currRow = worksheet.createRow(0);
for (int attribute = 0; attribute < metadataLL.size(); attribute++) {
Cell currCell = currRow.createCell(attribute);
currCell.setCellValue(metadataLL.get(attribute)[0]);
currCell.setCellStyle(attributeNameCells);
}
for (int node = 0; node < trainingDataLL.size(); node++) {
currRow = worksheet.createRow(node + 1); // Offset by one since first row is header data
int classifierCompleted = 0; // Necessary for if data does not end in classifier
for (int attribute = 0; attribute < metadataLL.size(); attribute++) {
Cell currCell = currRow.createCell(attribute);
if (metadataLL.get(attribute)[2].compareTo("classifier") == 0) {
currCell.setCellValue(knownClassifications.get(node));
classifierCompleted++;
} else {
currCell.setCellValue(trainingDataLL.get(node)[attribute - classifierCompleted]);
}
}
}
worksheet = workbook.createSheet("Likelihood");
currRow = worksheet.createRow(0);
int largestAttributeSize = 0;
for (int attribute = 0; attribute < classifier.size(); attribute++) {
if (classifier.get(attribute).size() > largestAttributeSize) {
largestAttributeSize = classifier.get(attribute).size();
}
}
// Label attributes along the top
for (int i = 0; i < metadataLL.size(); i++) {
if (i == 0) {
Cell currCell = currRow.createCell(i);
// currCell.setCellValue("Attributes");
currCell.setCellStyle(attributeNameCells);
} else {
Cell currCell = currRow.createCell(i);
currCell.setCellValue(
metadataLL
.get(i - 1)[0]); // -1 since the first cell does not contain a attribute name
currCell.setCellStyle(attributeNameCells);
}
}
// List possible classifications on the side and classification likelihoods at the end
for (int i = 0;
i
< (largestAttributeSize * (classificationTypes.size() + 1)
+ classificationTypes.size()
+ 1);
i++) // +1 since the first row of each stride lists each attributes string of what
// occurrence the likelihoods are displaying
{ // +classificationTypes.size() so we can list the classification types likelihood at the end
currRow = worksheet.createRow(i + 1); // +1 since first row is attribute names
Cell currCell = currRow.createCell(0);
int currentClassificationType =
i
% (classificationTypes.size()
+ 1); // +1 since the first row of each stride lists each attributes string of
// what occurrence the likelihoods are displaying
// List the classification type of each row along the side
if (i
< largestAttributeSize
* (classificationTypes.size()
+ 1)) // +1 since the first row of each stride lists each attributes string of
// what occurrence the likelihoods are displaying
{
for (int j = 0;
j < classificationTypes.size() + 1;
j++) // +1 since the first row of each stride lists each attributes string of what
// occurrence the likelihoods are displaying
{
if (currentClassificationType == 0) {
// Do nothing for now may have it say something later
} else if (currentClassificationType == j) {
currCell.setCellValue(
classificationTypes.get(
j - 1)); // -1 since the first cell does not contain a classification type
}
}
} else // List the likelihood of each classification at the end
{
for (int j = 0;
j < classificationTypes.size() + 1;
j++) // +1 since the first row of each stride lists each attributes string of what
// occurrence the likelihoods are displaying
{
if (currentClassificationType == 0) {
// Do nothing for now may have it say value later
} else if (currentClassificationType == j) {
currCell.setCellValue(
"Likelihood of: "
+ classificationTypes.get(j - 1)
+ " is "
+ classificationLikelihood.get(
j - 1)); // -1 since the first cell does not contain a classification type
}
}
}
currCell.setCellStyle(attributeNameCells);
}
// List the data
for (int attribute = 0; attribute < classifier.size(); attribute++) {
for (int occurrences = 0; occurrences < classifier.get(attribute).size(); occurrences++) {
for (int classification = 0;
classification < classifier.get(attribute).get(occurrences).length;
classification++) {
currRow =
worksheet.getRow(
(occurrences * classifier.get(attribute).get(occurrences).length
+ classification)
+ 1); // +1 since first row is attribute names
Cell currCell =
currRow.createCell(
(attribute) + 1); // TODO figure out why this errors out at i:0 j:4 k:0
// largestAttributeSize:105 on kidney dataset
currCell.setCellValue(classifier.get(attribute).get(occurrences)[classification]);
}
}
}
workbook.write(fileOut);
workbook.close();
workbook.close();
} catch (FileNotFoundException e) {
System.out.println("Error file not found");
e.printStackTrace();
System.exit(0);
} catch (IOException e) {
System.out.println("Unable to output file, is the output destination writelocked?");
e.printStackTrace();
System.exit(0);
}
}
public static void readExcelFile(String fileName) {
FileInputStream file;
try {
file = new FileInputStream(new File(fileName));
Workbook excelFile = new HSSFWorkbook(file);
Sheet sheet1 = excelFile.getSheetAt(0); // Data sheet
// Set just in case metadata is incomplete or malformed
classificationLocation =
sheet1.getRow(0).getPhysicalNumberOfCells()
- 1; // Minus one since classificationLocation includes 0 and getPhysicalNumberOfCells
// does not
Sheet sheet2 = excelFile.getSheetAt(1); // Metadata sheet
// Loop based on number of attribute names
for (int i = 0; i < sheet2.getRow(0).getPhysicalNumberOfCells(); i++) {
String[] metadata = new String[METADATASIZE];
// Construct metadata
Row currRow = sheet2.getRow(0); // This should be a row of names
metadata[0] = currRow.getCell(i).toString();
currRow = sheet2.getRow(1); // This should be a row of data types (discrete or continuous)
metadata[1] = currRow.getCell(i).toString();
currRow = sheet2.getRow(2); // This should say which one is the classifier
if (currRow.getCell(i) == null
|| currRow.getCell(i).getCellType() == Cell.CELL_TYPE_BLANK) {
metadata[2] = "attribute";
} else {
metadata[2] = "classifier";
classificationLocation = i;
}
metadataLL.add(metadata);
}
for (Row row : sheet1) {
String data[] = new String[row.getPhysicalNumberOfCells() - 1];
int offset =
0; // Used so that we can declare an array of the size of the attributes without the
// classification
for (Cell cell : row) {
int index = cell.getColumnIndex();
if (classificationLocation != index) {
data[index - offset] = cell.toString();
} else {
classificationsLL.add(cell.toString());
// Moved to generate training data so that we do not get possible classifications from
// unknown data since some denote unknown by saying ?
// //Check to see if we have seen it yet
//
// int occurrences = 0;
// for(int i = 0; i < classificationTypes.size(); i++)
// {
// if(classificationTypes.get(i).compareTo(cell.toString()) == 0)
// {
// occurrences++;
// }
// }
// if(occurrences == 0)
// {
// classificationTypes.add(cell.toString());
// }
offset++;
}
}
dataLL.add(data);
// classCount = temp.length;
}
excelFile.close();
} catch (FileNotFoundException e) {
System.out.println("Error file not found");
System.exit(0);
} catch (IOException e) {
System.out.println("Unable to read file, disk drive may be failing");
e.printStackTrace();
System.exit(0);
}
}