Adapt enum type params in pca, chi square test, glm and correlation. …

…see #73
alibaba · Apr 9, 2020 · 309ecb6 · 309ecb6
1 parent 5a090a3
commit 309ecb6
Show file tree

Hide file tree

Showing 23 changed files with 205 additions and 171 deletions.
diff --git a/core/src/main/java/com/alibaba/alink/operator/batch/feature/ChiSqSelectorBatchOp.java b/core/src/main/java/com/alibaba/alink/operator/batch/feature/ChiSqSelectorBatchOp.java
@@ -25,7 +25,7 @@ public ChiSqSelectorBatchOp linkFrom(BatchOperator<?>... inputs) {
         String[] selectedColNames = getSelectedCols();
         String labelColName = getLabelCol();
 
-        String selectorType = getParams().get(SELECTOR_TYPE).trim().toLowerCase();
+        SelectorType selectorType = getParams().get(SELECTOR_TYPE);
         int numTopFeatures = getParams().get(NUM_TOP_FEATURES);
         double percentile = getParams().get(PERCENTILE);
         double fpr = getParams().get(FPR);

diff --git a/core/src/main/java/com/alibaba/alink/operator/batch/feature/PcaTrainBatchOp.java b/core/src/main/java/com/alibaba/alink/operator/batch/feature/PcaTrainBatchOp.java
@@ -1,7 +1,6 @@
 package com.alibaba.alink.operator.batch.feature;
 
 import com.alibaba.alink.common.linalg.*;
-import com.alibaba.alink.operator.common.feature.pca.PcaTypeEnum;
 import org.apache.flink.api.common.functions.RichFlatMapFunction;
 import org.apache.flink.api.common.functions.RichMapPartitionFunction;
 import org.apache.flink.api.java.DataSet;
@@ -328,7 +327,7 @@ public void mapPartition(Iterable<Tuple2<Integer, DenseVector>> splitVec, Collec
             PcaModelData pcr = new PcaModelData();
 
             //get correlation or covariance matrix
-            PcaTypeEnum pcaTypeEnum = PcaTypeEnum.valueOf(pcaType.toUpperCase());
+            CalculationType pcaTypeEnum = CalculationType.valueOf(pcaType.toUpperCase());
 
             double[][] corr = null;
 
@@ -346,7 +345,7 @@ public void mapPartition(Iterable<Tuple2<Integer, DenseVector>> splitVec, Collec
 
 
             DenseMatrix calculateMatrix = new DenseMatrix(corr);
-            if (pcaTypeEnum.equals(PcaTypeEnum.COVAR_POP)) {
+            if (pcaTypeEnum.equals(CalculationType.COVAR_POP)) {
                 double cnt = counts[0];
                 if (cnt > 1) {
                     calculateMatrix.scaleEqual(cnt / (cnt - 1));

diff --git a/core/src/main/java/com/alibaba/alink/operator/batch/feature/VectorChiSqSelectorBatchOp.java b/core/src/main/java/com/alibaba/alink/operator/batch/feature/VectorChiSqSelectorBatchOp.java
@@ -24,7 +24,7 @@ public VectorChiSqSelectorBatchOp linkFrom(BatchOperator<?>... inputs) {
         String vectorColName = getSelectedCol();
         String labelColName = getLabelCol();
 
-        String selectorType = getParams().get(SELECTOR_TYPE).trim().toLowerCase();
+        SelectorType selectorType = getParams().get(SELECTOR_TYPE);
         int numTopFeatures = getParams().get(NUM_TOP_FEATURES);
         double percentile = getParams().get(PERCENTILE);
         double fpr = getParams().get(FPR);

diff --git a/core/src/main/java/com/alibaba/alink/operator/batch/regression/GlmEvaluationBatchOp.java b/core/src/main/java/com/alibaba/alink/operator/batch/regression/GlmEvaluationBatchOp.java
@@ -36,8 +36,8 @@ public GlmEvaluationBatchOp linkFrom(BatchOperator<?>... inputs) {
         String weightColName = getWeightCol();
         String offsetColName = getOffsetCol();
 
-        String familyName = getFamily();
-        String linkName = getLink();
+        Family familyName = getFamily();
+        Link linkName = getLink();
         double variancePower = getVariancePower();
         double linkPower = getLinkPower();
 

diff --git a/core/src/main/java/com/alibaba/alink/operator/batch/regression/GlmTrainBatchOp.java b/core/src/main/java/com/alibaba/alink/operator/batch/regression/GlmTrainBatchOp.java
@@ -48,8 +48,8 @@ public GlmTrainBatchOp linkFrom(BatchOperator<?>... inputs) {
         String weightColName = getWeightCol();
         String offsetColName = getOffsetCol();
 
-        String familyName = getFamily();
-        String linkName = getLink();
+        Family familyName = getFamily();
+        Link linkName = getLink();
         double variancePower = getVariancePower();
         double linkPower = getLinkPower();
 
@@ -125,9 +125,9 @@ private static class BuildModel implements MapPartitionFunction<GlmUtil.Weighted
         private String weightColName;
         private String labelColName;
 
-        private String familyName;
+        private Family familyName;
         private double variancePower;
-        private String linkName;
+        private Link linkName;
         private double linkPower;
 
         private boolean fitIntercept;
@@ -136,8 +136,8 @@ private static class BuildModel implements MapPartitionFunction<GlmUtil.Weighted
 
         public BuildModel(String[] featureColNames, String offsetColName,
                           String weightColName, String labelColName,
-                          String familyName, double variancePower,
-                          String linkName, double linkPower,
+                          Family familyName, double variancePower,
+                          Link linkName, double linkPower,
                           Boolean fitIntercept, int numIter, double epsilon) {
             this.featureColNames = featureColNames;
             this.offsetColName = offsetColName;

diff --git a/core/src/main/java/com/alibaba/alink/operator/batch/statistics/CorrelationBatchOp.java b/core/src/main/java/com/alibaba/alink/operator/batch/statistics/CorrelationBatchOp.java
@@ -47,9 +47,9 @@ public CorrelationBatchOp linkFrom(BatchOperator<?>... inputs) {
         //check col types must be double or bigint
         TableUtil.assertNumericalCols(in.getSchema(), selectedColNames);
 
-        String corrType = getMethod().trim().toLowerCase();
+        Method corrType = getMethod();
 
-        if ("pearson".equals(corrType)) {
+        if (Method.PEARSON == corrType) {
 
             DataSet<Tuple2<TableSummary, CorrelationResult>> srt = StatisticsHelper.pearsonCorrelation(in, selectedColNames);
 

diff --git a/core/src/main/java/com/alibaba/alink/operator/batch/statistics/VectorCorrelationBatchOp.java b/core/src/main/java/com/alibaba/alink/operator/batch/statistics/VectorCorrelationBatchOp.java
@@ -39,9 +39,9 @@ public VectorCorrelationBatchOp linkFrom(BatchOperator<?>... inputs) {
         BatchOperator<?> in = checkAndGetFirst(inputs);
         String vectorColName = getSelectedCol();
 
-        String corrType = getMethod().trim().toLowerCase();
+        Method corrType = getMethod();
 
-        if ("pearson".equals(corrType)) {
+        if (Method.PEARSON == corrType) {
             DataSet<Tuple2<BaseVectorSummary, CorrelationResult>> srt = StatisticsHelper.vectorPearsonCorrelation(in, vectorColName);
 
             //block

diff --git a/core/src/main/java/com/alibaba/alink/operator/common/feature/pca/PcaModelDataConverter.java b/core/src/main/java/com/alibaba/alink/operator/common/feature/pca/PcaModelDataConverter.java
@@ -7,6 +7,7 @@
 import org.apache.flink.ml.api.misc.param.Params;
 
 import com.alibaba.alink.common.utils.JsonConverter;
+import com.alibaba.alink.params.feature.HasCalculationType;
 import com.alibaba.alink.params.shared.colname.HasFeatureColsDefaultAsNull;
 import com.alibaba.alink.params.feature.PcaTrainParams;
 

diff --git a/core/src/main/java/com/alibaba/alink/operator/common/feature/pca/PcaModelMapper.java b/core/src/main/java/com/alibaba/alink/operator/common/feature/pca/PcaModelMapper.java
@@ -6,6 +6,7 @@
 import com.alibaba.alink.common.mapper.ModelMapper;
 import com.alibaba.alink.common.utils.OutputColsHelper;
 import com.alibaba.alink.common.utils.TableUtil;
+import com.alibaba.alink.params.feature.HasCalculationType;
 import com.alibaba.alink.params.feature.PcaPredictParams;
 import org.apache.flink.ml.api.misc.param.Params;
 import org.apache.flink.table.api.TableSchema;
@@ -25,7 +26,7 @@ public class PcaModelMapper extends ModelMapper {
     private int[] featureIdxs = null;
     private boolean isVector;
 
-    private String transformType = null;
+    private PcaPredictParams.TransformType transformType = null;
     private String pcaType = null;
 
     private double[] sourceMean = null;
@@ -77,8 +78,7 @@ public void loadModel(List<Row> modelRows) {
         int nx = model.means.length;
         int p = model.p;
 
-        PcaTransformTypeEnum transformTypeEnum = PcaTransformTypeEnum.valueOf(this.transformType.toUpperCase());
-        PcaTypeEnum pcaTypeEnum = PcaTypeEnum.valueOf(this.pcaType.toUpperCase());
+        HasCalculationType.CalculationType pcaTypeEnum = HasCalculationType.CalculationType.valueOf(this.pcaType.toUpperCase());
 
         //transform mean, stdDevs and scoreStd
         sourceMean = new double[nx];
@@ -88,11 +88,11 @@ public void loadModel(List<Row> modelRows) {
         Arrays.fill(sourceStd, 1);
         Arrays.fill(scoreStd, 1);
 
-        if (PcaTypeEnum.CORR.equals(pcaTypeEnum)) {
+        if (HasCalculationType.CalculationType.CORR.equals(pcaTypeEnum)) {
             sourceStd = model.stddevs;
         }
 
-        switch (transformTypeEnum) {
+        switch (transformType) {
             case SUBMEAN:
                 sourceMean = model.means;
                 break;
@@ -108,6 +108,10 @@ public void loadModel(List<Row> modelRows) {
                     scoreStd[i] = Math.sqrt(tmp);
                 }
                 break;
+            case SIMPLE:
+                break;
+            default:
+                throw new IllegalArgumentException("Error transformType: " + transformType);
         }
     }
 

diff --git a/core/src/main/java/com/alibaba/alink/operator/common/feature/pca/PcaTransformTypeEnum.java b/core/src/main/java/com/alibaba/alink/operator/common/feature/pca/PcaTransformTypeEnum.java
diff --git a/core/src/main/java/com/alibaba/alink/operator/common/feature/pca/PcaTypeEnum.java b/core/src/main/java/com/alibaba/alink/operator/common/feature/pca/PcaTypeEnum.java
diff --git a/core/src/main/java/com/alibaba/alink/operator/common/regression/GlmModelData.java b/core/src/main/java/com/alibaba/alink/operator/common/regression/GlmModelData.java
@@ -1,6 +1,8 @@
 package com.alibaba.alink.operator.common.regression;
 
 
+import com.alibaba.alink.params.regression.GlmTrainParams;
+
 /**
  * Glm model data.
  */
@@ -51,7 +53,7 @@ public class GlmModelData {
     /**
      * family name.
      */
-    public String familyName;
+    public GlmTrainParams.Family familyName;
 
     /**
      * variance power of family.
@@ -61,7 +63,7 @@ public class GlmModelData {
     /**
      * link function name.
      */
-    public String linkName;
+    public GlmTrainParams.Link linkName;
 
     /**
      * power of link function.

diff --git a/core/src/main/java/com/alibaba/alink/operator/common/regression/GlmModelMapper.java b/core/src/main/java/com/alibaba/alink/operator/common/regression/GlmModelMapper.java
@@ -98,9 +98,9 @@ public void loadModel(List<Row> modelRows) {
 
         features = new double[featureColIdxs.length];
 
-        String familyName = params.get(GlmTrainParams.FAMILY);
+        GlmTrainParams.Family familyName = params.get(GlmTrainParams.FAMILY);
         double variancePower = params.get(GlmTrainParams.VARIANCE_POWER);
-        String linkName = params.get(GlmTrainParams.LINK);
+        GlmTrainParams.Link linkName = params.get(GlmTrainParams.LINK);
         double linkPower = params.get(GlmTrainParams.LINK_POWER);
 
         familyLink = new FamilyLink(familyName, variancePower, linkName, linkPower);

diff --git a/core/src/main/java/com/alibaba/alink/operator/common/regression/glm/FamilyLink.java b/core/src/main/java/com/alibaba/alink/operator/common/regression/glm/FamilyLink.java
@@ -15,6 +15,7 @@
 import com.alibaba.alink.operator.common.regression.glm.link.Power;
 import com.alibaba.alink.operator.common.regression.glm.link.Probit;
 import com.alibaba.alink.operator.common.regression.glm.link.Sqrt;
+import com.alibaba.alink.params.regression.GlmTrainParams;
 
 import java.io.Serializable;
 
@@ -32,57 +33,57 @@ public class FamilyLink implements Serializable {
      * @param linkName: link name.
      * @param linkPower: link power.
      */
-    public FamilyLink(String familyName, double variancePower, String linkName, double linkPower) {
-        if (familyName == null || familyName.isEmpty()) {
+    public FamilyLink(GlmTrainParams.Family familyName, double variancePower, GlmTrainParams.Link linkName, double linkPower) {
+        if (familyName == null) {
             throw new RuntimeException("family can not be empty");
         }
 
-        switch (familyName.toLowerCase()) {
-            case "gamma":
+        switch (familyName) {
+            case Gamma:
                 family = new Gamma();
                 break;
-            case "binomial":
+            case Binomial:
                 family = new Binomial();
                 break;
-            case "gaussian":
+            case Gaussian:
                 family = new Gaussian();
                 break;
-            case "poisson":
+            case Poisson:
                 family = new Poisson();
                 break;
-            case "tweedie":
+            case Tweedie:
                 family = new Tweedie(variancePower);
                 break;
             default:
                 throw new RuntimeException("family is not support. ");
         }
 
-        if (linkName == null || linkName.isEmpty()) {
+        if (linkName == null) {
             link = family.getDefaultLink();
         } else {
-            switch (linkName.toLowerCase()) {
-                case "cloglog":
+            switch (linkName) {
+                case CLogLog:
                     link = new CLogLog();
                     break;
-                case "identity":
+                case Identity:
                     link = new Identity();
                     break;
-                case "inverse":
+                case Inverse:
                     link = new Inverse();
                     break;
-                case "log":
+                case Log:
                     link = new Log();
                     break;
-                case "logit":
+                case Logit:
                     link = new Logit();
                     break;
-                case "power":
+                case Power:
                     link = new Power(linkPower);
                     break;
-                case "probit":
+                case Probit:
                     link = new Probit();
                     break;
-                case "sqrt":
+                case Sqrt:
                     link = new Sqrt();
                     break;
                 default: