137 lines
5.3 KiB
Java
137 lines
5.3 KiB
Java
package eu.dnetlib.jobs.featureextraction.lda;
|
|
|
|
import com.clearspring.analytics.util.Lists;
|
|
import eu.dnetlib.featureextraction.Utilities;
|
|
import eu.dnetlib.jobs.AbstractSparkJob;
|
|
import eu.dnetlib.support.ArgumentApplicationParser;
|
|
import eu.dnetlib.support.Author;
|
|
import eu.dnetlib.support.AuthorsFactory;
|
|
import org.apache.spark.SparkConf;
|
|
import org.apache.spark.api.java.JavaDoubleRDD;
|
|
import org.apache.spark.api.java.JavaRDD;
|
|
import org.apache.spark.api.java.JavaSparkContext;
|
|
import org.apache.spark.sql.SparkSession;
|
|
import org.codehaus.jackson.map.ObjectMapper;
|
|
import org.slf4j.Logger;
|
|
import org.slf4j.LoggerFactory;
|
|
import scala.Tuple2;
|
|
|
|
import java.io.IOException;
|
|
import java.util.ArrayList;
|
|
import java.util.Iterator;
|
|
import java.util.List;
|
|
import java.util.Optional;
|
|
import java.util.stream.Collectors;
|
|
|
|
public class SparkLDAAnalysis extends AbstractSparkJob {
|
|
|
|
private static final Logger log = LoggerFactory.getLogger(SparkLDAAnalysis.class);
|
|
|
|
public SparkLDAAnalysis(ArgumentApplicationParser parser, SparkSession spark) {
|
|
super(parser, spark);
|
|
}
|
|
public static void main(String[] args) throws Exception {
|
|
ArgumentApplicationParser parser = new ArgumentApplicationParser(
|
|
readResource("/jobs/parameters/ldaAnalysis_parameters.json", SparkLDATuning.class)
|
|
);
|
|
|
|
parser.parseArgument(args);
|
|
|
|
SparkConf conf = new SparkConf();
|
|
|
|
new SparkLDAAnalysis(
|
|
parser,
|
|
getSparkSession(conf)
|
|
).run();
|
|
}
|
|
|
|
@Override
|
|
protected void run() throws IOException {
|
|
// read oozie parameters
|
|
final String authorsPath = parser.get("authorsPath");
|
|
final String workingPath = parser.get("workingPath");
|
|
final int numPartitions = Optional
|
|
.ofNullable(parser.get("numPartitions"))
|
|
.map(Integer::valueOf)
|
|
.orElse(NUM_PARTITIONS);
|
|
|
|
log.info("authorsPath: '{}'", authorsPath);
|
|
log.info("workingPath: '{}'", workingPath);
|
|
log.info("numPartitions: '{}'", numPartitions);
|
|
|
|
JavaSparkContext context = JavaSparkContext.fromSparkContext(spark.sparkContext());
|
|
|
|
JavaRDD<Author> authors = context
|
|
.textFile(authorsPath)
|
|
.map(s -> new ObjectMapper().readValue(s, Author.class))
|
|
.filter(a -> !a.getOrcid().isEmpty()); //don't need authors without orcid for the threshold analysis
|
|
|
|
JavaRDD<Tuple2<Boolean, Double>> groundTruthThreshold = authors
|
|
.mapToPair(a -> new Tuple2<>(AuthorsFactory.getLNFI(a), a))
|
|
.flatMapToPair(a -> a._1().stream().map(k -> new Tuple2<>(k, a._2())).collect(Collectors.toList()).iterator())
|
|
.groupByKey()
|
|
.flatMap(a -> thresholdAnalysis(a._2()));
|
|
|
|
JavaDoubleRDD groundTruthTrue = groundTruthThreshold.filter(Tuple2::_1).mapToDouble(Tuple2::_2);
|
|
long totalPositives = groundTruthTrue.count();
|
|
JavaDoubleRDD groundTruthFalse = groundTruthThreshold.filter(x -> !x._1()).mapToDouble(Tuple2::_2);
|
|
long totalNegatives = groundTruthFalse.count();
|
|
|
|
double[] thresholds = new double[]{0.0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0};
|
|
|
|
List<String> stats = new ArrayList<>();
|
|
stats.add("th,fp,fn,tp,tn,total_positives,total_negatives");
|
|
for(double threshold: thresholds) {
|
|
long truePositive = groundTruthTrue.filter(d -> d >= threshold).count();
|
|
long falsePositive = groundTruthFalse.filter(d -> d >= threshold).count();
|
|
long trueNegative = groundTruthFalse.filter(d -> d < threshold).count();
|
|
long falseNegative = groundTruthTrue.filter(d -> d < threshold).count();
|
|
|
|
stats.add(threshold + "," + falsePositive + "," + falseNegative + "," + truePositive + "," + trueNegative + "," + totalPositives + "," + totalNegatives);
|
|
}
|
|
|
|
Utilities.writeLinesToHDFSFile(stats, workingPath + "/threshold_analysis.csv");
|
|
|
|
}
|
|
|
|
public Iterator<Tuple2<Boolean, Double>> thresholdAnalysis(Iterable<Author> a) {
|
|
List<Author> authors = Lists.newArrayList(a);
|
|
|
|
List<Tuple2<Boolean, Double>> results = new ArrayList<>();
|
|
int i = 0;
|
|
int j = 1;
|
|
while(i < authors.size()) {
|
|
|
|
while(j < authors.size()) {
|
|
boolean bRes;
|
|
if(authors.get(i).getOrcid().isEmpty() || authors.get(j).getOrcid().isEmpty())
|
|
bRes = false;
|
|
else {
|
|
bRes = authors.get(i).getOrcid().equals(authors.get(j).getOrcid());
|
|
}
|
|
results.add(new Tuple2<>(bRes, cosineSimilarity(authors.get(i).getEmbeddings().get("lda_topics"), authors.get(j).getEmbeddings().get("lda_topics"))));
|
|
j++;
|
|
}
|
|
i++;
|
|
j=i+1;
|
|
}
|
|
|
|
return results.iterator();
|
|
}
|
|
|
|
double cosineSimilarity(double[] a, double[] b) {
|
|
double dotProduct = 0;
|
|
double normASum = 0;
|
|
double normBSum = 0;
|
|
|
|
for(int i = 0; i < a.length; i ++) {
|
|
dotProduct += a[i] * b[i];
|
|
normASum += a[i] * a[i];
|
|
normBSum += b[i] * b[i];
|
|
}
|
|
|
|
double eucledianDist = Math.sqrt(normASum) * Math.sqrt(normBSum);
|
|
return dotProduct / eucledianDist;
|
|
}
|
|
}
|