Java KMeansData类代码示例

import org.apache.flink.examples.java.clustering.util.KMeansData; //导入依赖的package包/类
private static void runKMeans(ExecutionEnvironment env) throws Exception {

		env.setParallelism(PARALLELISM);
		env.getConfig().disableSysoutLogging();

		// get input data
		DataSet<KMeans.Point> points =  KMeansData.getDefaultPointDataSet(env).rebalance();
		DataSet<KMeans.Centroid> centroids =  KMeansData.getDefaultCentroidDataSet(env).rebalance();

		// set number of bulk iterations for KMeans algorithm
		IterativeDataSet<KMeans.Centroid> loop = centroids.iterate(20);

		DataSet<KMeans.Centroid> newCentroids = points
				// compute closest centroid for each point
				.map(new KMeans.SelectNearestCenter()).withBroadcastSet(loop, "centroids")
						// count and sum point coordinates for each centroid
				.map(new KMeans.CountAppender())
				.groupBy(0).reduce(new KMeans.CentroidAccumulator())
						// compute new centroids from point counts and coordinate sums
				.map(new KMeans.CentroidAverager());

		// feed new centroids back into next iteration
		DataSet<KMeans.Centroid> finalCentroids = loop.closeWith(newCentroids);

		DataSet<Tuple2<Integer, KMeans.Point>> clusteredPoints = points
				// assign points to final clusters
				.map(new KMeans.SelectNearestCenter()).withBroadcastSet(finalCentroids, "centroids");

		clusteredPoints.output(new DiscardingOutputFormat<Tuple2<Integer, KMeans.Point>>());

		env.execute("KMeans Example");
	}
 

import org.apache.flink.examples.java.clustering.util.KMeansData; //导入依赖的package包/类
private static DataSet<Centroid> getCentroidDataSet(ParameterTool params, ExecutionEnvironment env) {
	DataSet<Centroid> centroids;
	if (params.has("centroids")) {
		centroids = env.readCsvFile(params.get("centroids"))
			.fieldDelimiter(" ")
			.pojoType(Centroid.class, "id", "x", "y");
	} else {
		System.out.println("Executing K-Means example with default centroid data set.");
		System.out.println("Use --centroids to specify file input.");
		centroids = KMeansData.getDefaultCentroidDataSet(env);
	}
	return centroids;
}
 

import org.apache.flink.examples.java.clustering.util.KMeansData; //导入依赖的package包/类
private static DataSet<Point> getPointDataSet(ParameterTool params, ExecutionEnvironment env) {
	DataSet<Point> points;
	if (params.has("points")) {
		// read points from CSV file
		points = env.readCsvFile(params.get("points"))
			.fieldDelimiter(" ")
			.pojoType(Point.class, "x", "y");
	} else {
		System.out.println("Executing K-Means example with default point data set.");
		System.out.println("Use --points to specify file input.");
		points = KMeansData.getDefaultPointDataSet(env);
	}
	return points;
}
 

import org.apache.flink.examples.java.clustering.util.KMeansData; //导入依赖的package包/类
private static DataSet<Point> getPointDataSet(ExecutionEnvironment env) {
	if(fileOutput) {
		// read points from CSV file
		return env.readCsvFile(pointsPath)
					.fieldDelimiter(' ')
					.includeFields(true, true)
					.types(Double.class, Double.class)
					.map(new TuplePointConverter());
	} else {
		return KMeansData.getDefaultPointDataSet(env);
	}
}
 

import org.apache.flink.examples.java.clustering.util.KMeansData; //导入依赖的package包/类
private static DataSet<Centroid> getCentroidDataSet(ExecutionEnvironment env) {
	if(fileOutput) {
		return env.readCsvFile(centersPath)
					.fieldDelimiter(' ')
					.includeFields(true, true, true)
					.types(Integer.class, Double.class, Double.class)
					.map(new TupleCentroidConverter());
	} else {
		return KMeansData.getDefaultCentroidDataSet(env);
	}
}