Examples with PCA com.alibaba.alink.pipeline.feature.PCA used on opensource projects

Example 1 with PCA

use of com.alibaba.alink.pipeline.feature.PCA in project Alink by alibaba.

the class Chap19 method c_2.

static void c_2() throws Exception {
    MemSourceBatchOp source = new MemSourceBatchOp(CRIME_ROWS_DATA, CRIME_COL_NAMES);
    Pipeline std_pca = new Pipeline().add(new StandardScaler().setSelectedCols("murder", "rape", "robbery", "assault", "burglary", "larceny", "auto")).add(new PCA().setCalculationType(CalculationType.COV).setK(4).setSelectedCols("murder", "rape", "robbery", "assault", "burglary", "larceny", "auto").setPredictionCol(VECTOR_COL_NAME).enableLazyPrintModelInfo());
    std_pca.fit(source).transform(source).link(new VectorToColumnsBatchOp().setVectorCol(VECTOR_COL_NAME).setSchemaStr("prin1 double, prin2 double, prin3 double, prin4 double").setReservedCols("state")).lazyPrint(10, "state with principle components");
    BatchOperator.execute();
}

Example 2 with PCA

use of com.alibaba.alink.pipeline.feature.PCA in project Alink by alibaba.

the class Chap19 method c_4.

static void c_4() throws Exception {
    AkSourceBatchOp dense_train_data = new AkSourceBatchOp().setFilePath(DATA_DIR + DENSE_TRAIN_FILE);
    AkSourceBatchOp dense_test_data = new AkSourceBatchOp().setFilePath(DATA_DIR + DENSE_TEST_FILE);
    AkSourceBatchOp sparse_train_data = new AkSourceBatchOp().setFilePath(DATA_DIR + SPARSE_TRAIN_FILE);
    AkSourceBatchOp sparse_test_data = new AkSourceBatchOp().setFilePath(DATA_DIR + SPARSE_TEST_FILE);
    Stopwatch sw = new Stopwatch();
    sw.reset();
    sw.start();
    new KnnClassifier().setK(3).setVectorCol(VECTOR_COL_NAME).setLabelCol(LABEL_COL_NAME).setPredictionCol(PREDICTION_COL_NAME).fit(dense_train_data).transform(dense_test_data).link(new EvalMultiClassBatchOp().setLabelCol(LABEL_COL_NAME).setPredictionCol(PREDICTION_COL_NAME).lazyPrintMetrics("KnnClassifier Dense"));
    BatchOperator.execute();
    sw.stop();
    System.out.println(sw.getElapsedTimeSpan());
    sw.reset();
    sw.start();
    new KnnClassifier().setK(3).setVectorCol(VECTOR_COL_NAME).setLabelCol(LABEL_COL_NAME).setPredictionCol(PREDICTION_COL_NAME).fit(sparse_train_data).transform(sparse_test_data).link(new EvalMultiClassBatchOp().setLabelCol(LABEL_COL_NAME).setPredictionCol(PREDICTION_COL_NAME).lazyPrintMetrics("KnnClassifier Sparse"));
    BatchOperator.execute();
    sw.stop();
    System.out.println(sw.getElapsedTimeSpan());
    sw.reset();
    sw.start();
    new Pipeline().add(new PCA().setK(39).setCalculationType(CalculationType.COV).setVectorCol(VECTOR_COL_NAME).setPredictionCol(VECTOR_COL_NAME)).add(new KnnClassifier().setK(3).setVectorCol(VECTOR_COL_NAME).setLabelCol(LABEL_COL_NAME).setPredictionCol(PREDICTION_COL_NAME)).fit(dense_train_data).transform(dense_test_data).link(new EvalMultiClassBatchOp().setLabelCol(LABEL_COL_NAME).setPredictionCol(PREDICTION_COL_NAME).lazyPrintMetrics("Knn with PCA Dense"));
    BatchOperator.execute();
    sw.stop();
    System.out.println(sw.getElapsedTimeSpan());
    sw.reset();
    sw.start();
    new Pipeline().add(new PCA().setK(39).setCalculationType(CalculationType.COV).setVectorCol(VECTOR_COL_NAME).setPredictionCol(VECTOR_COL_NAME)).add(new KnnClassifier().setK(3).setVectorCol(VECTOR_COL_NAME).setLabelCol(LABEL_COL_NAME).setPredictionCol(PREDICTION_COL_NAME)).fit(sparse_train_data).transform(sparse_test_data).link(new EvalMultiClassBatchOp().setLabelCol(LABEL_COL_NAME).setPredictionCol(PREDICTION_COL_NAME).lazyPrintMetrics("Knn with PCA Sparse"));
    BatchOperator.execute();
    sw.stop();
    System.out.println(sw.getElapsedTimeSpan());
    sw.reset();
    sw.start();
    new Pipeline().add(new PCAModel().setVectorCol(VECTOR_COL_NAME).setPredictionCol(VECTOR_COL_NAME).setModelData(new AkSourceBatchOp().setFilePath(DATA_DIR + PCA_MODEL_FILE))).add(new KnnClassifier().setK(3).setVectorCol(VECTOR_COL_NAME).setLabelCol(LABEL_COL_NAME).setPredictionCol(PREDICTION_COL_NAME)).fit(dense_train_data).transform(dense_test_data).link(new EvalMultiClassBatchOp().setLabelCol(LABEL_COL_NAME).setPredictionCol(PREDICTION_COL_NAME).lazyPrintMetrics("Knn PCAModel Dense"));
    BatchOperator.execute();
    sw.stop();
    System.out.println(sw.getElapsedTimeSpan());
    sw.reset();
    sw.start();
    new Pipeline().add(new PCAModel().setVectorCol(VECTOR_COL_NAME).setPredictionCol(VECTOR_COL_NAME).setModelData(new AkSourceBatchOp().setFilePath(DATA_DIR + PCA_MODEL_FILE))).add(new KnnClassifier().setK(3).setVectorCol(VECTOR_COL_NAME).setLabelCol(LABEL_COL_NAME).setPredictionCol(PREDICTION_COL_NAME)).fit(sparse_train_data).transform(sparse_test_data).link(new EvalMultiClassBatchOp().setLabelCol(LABEL_COL_NAME).setPredictionCol(PREDICTION_COL_NAME).lazyPrintMetrics("Knn PCAModel Sparse"));
    BatchOperator.execute();
    sw.stop();
    System.out.println(sw.getElapsedTimeSpan());
}

Example 3 with PCA

use of com.alibaba.alink.pipeline.feature.PCA in project Alink by alibaba.

the class Chap19 method c_1.

static void c_1() throws Exception {
    MemSourceBatchOp source = new MemSourceBatchOp(CRIME_ROWS_DATA, CRIME_COL_NAMES);
    source.lazyPrint(10, "Origin data");
    BatchOperator<?> pca_result = new PCA().setK(4).setSelectedCols("murder", "rape", "robbery", "assault", "burglary", "larceny", "auto").setPredictionCol(VECTOR_COL_NAME).enableLazyPrintModelInfo().fit(source).transform(source).link(new VectorToColumnsBatchOp().setVectorCol(VECTOR_COL_NAME).setSchemaStr("prin1 double, prin2 double, prin3 double, prin4 double").setReservedCols("state")).lazyPrint(10, "state with principle components");
    pca_result.select("state, prin1").orderBy("prin1", 100, false).lazyPrint(-1, "Order by prin1");
    pca_result.select("state, prin2").orderBy("prin2", 100, false).lazyPrint(-1, "Order by prin2");
    BatchOperator.execute();
}