@SuppressWarnings("unchecked")
public double crossValidationError(int n, Approach approach) throws IOException {
List<String> queries = QueryLoader.loadQueries(queryFile);
int partSize = queries.size() / n;
List<List<String>> partitions = new ArrayList<>(n);
for (int i = 0; i < (n - 1); i++) {
partitions.add(queries.subList(i * partSize, (i + 1) * partSize));
}
partitions.add(queries.subList((n - 1) * partSize, queries.size()));
Model model = readModel(MODEL_FILE);
LOGGER.info("Generating expected counts...");
ObjectIntOpenHashMap<String> gsResults[] = countResources(partitions, model);
double rootMeanSquareSum = 0;
double foldErrors[] = new double[n];
List<String> training, predicted;
for (int i = 0; i < n; i++) {
LOGGER.info("Starting fold " + i + "...");
training = generateTrainingSet(i, partitions);
predicted = approach.generateResourceRanking(training, model);
foldErrors[i] = RMSD.getRMSD(predicted, generateExpectedResult(i, gsResults));
LOGGER.info("Error of fold " + i + " = " + foldErrors[i]);
rootMeanSquareSum += foldErrors[i];
}
LOGGER.info("Error of folds " + Arrays.toString(foldErrors));
return rootMeanSquareSum / n;
}
public double validate(Approach approach, Model model, QueryExecutor executor)
throws IOException {
List<String> queries = QueryLoader.loadQueries(queryFile);
Baseline baseline = new Baseline(executor);
return RMSD.getRMSD(
approach.generateResourceRanking(queries, model),
generateUriRankRangeMapping(baseline.sumUpResults(model, queries)));
}