/**
* @param file is the file to be classified
* @return returns true if correctly classified otherwise false
*/
public boolean classify(String file) {
String category = "";
double category_prob = 0;
// Read document;
DocumentI document = new Document(vocab);
document.read_file(file, false);
// Mapping for document
Map<String, Integer> document_dictionary = document.get_words();
// Distinct words in vocabulary
Set<String> vocab_distinct_words = vocab.distinct_words();
for (Entry<String, TextI> entry : categories.entrySet()) {
double pv = (double) entry.getValue().num_docs() / (double) total_docs;
double pwv = 1;
double prevpwv = 1;
for (String word : document_dictionary.keySet()) {
if (vocab_distinct_words.contains(word)) {
if ((pwv *= entry.getValue().pwv(word)) == 0) {
pwv = prevpwv;
break;
}
prevpwv = pwv;
}
}
if (pwv * pv > category_prob) {
category = entry.getKey();
category_prob = pwv * pv;
}
}
String predicted_category = category.replace(this.directory + "\\", "");
String actual_category = document.getCategory();
if (actual_category.contains(predicted_category)) {
return true;
} else {
return false;
}
}
/**
* @param dir is 20_newsgroups directory
* @param vocab is the vocabulary of the naive bayes classifier Currently hard coded to accept
* 20_newsgroups folder where each sub-folder is treated as a category
*/
public void learn(String dir, VocabI vocab) {
this.directory = dir;
this.vocab = vocab;
File folder = new File(dir);
if (!folder.isDirectory()) {
System.out.println("Please provide directory");
return;
}
File[] subdirectories = folder.listFiles();
for (File subdirectory : subdirectories) {
TextI category_text = new Text(vocab);
if (subdirectory.isDirectory()) {
File[] files = subdirectory.listFiles();
for (File file : files) {
category_text.read_file(file.toString(), true);
total_docs++;
}
}
categories.put(subdirectory.toString(), category_text);
}
vocab.refine();
}
