Refactoring the scalers.

datumbox · Dec 27, 2016 · 4dc9911 · 4dc9911
1 parent eac571e
commit 4dc9911
Show file tree

Hide file tree

Showing 5 changed files with 76 additions and 72 deletions.
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -1,7 +1,7 @@
 CHANGELOG
 =========
 
-Version 0.8.0-SNAPSHOT - Build 20161226
+Version 0.8.0-SNAPSHOT - Build 20161227
 ---------------------------------------
 
 - Initial Updates:

diff --git a/README.md b/README.md
@@ -16,7 +16,7 @@ The code is licensed under the [Apache License, Version 2.0](./LICENSE).
 Version
 -------
 
-The latest version is 0.8.0-SNAPSHOT (Build 20161226).
+The latest version is 0.8.0-SNAPSHOT (Build 20161227).
 
 The [devel branch](https://github.com/datumbox/datumbox-framework/tree/devel) is the development branch (default github branch). The [master branch](https://github.com/datumbox/datumbox-framework/tree/master) contains the latest stable version of the framework. All the stable releases are marked with [tags](https://github.com/datumbox/datumbox-framework/releases).
 

diff --git a/...src/main/java/com/datumbox/framework/core/machinelearning/preprocessing/MaxAbsScaler.java b/...src/main/java/com/datumbox/framework/core/machinelearning/preprocessing/MaxAbsScaler.java
@@ -138,30 +138,15 @@ protected void _transform(Dataframe newData) {
 
                 Double maxAbsolute = entry.getValue();
 
-                double normalizedValue;
-                if(maxAbsolute.equals(0.0)) {
-                    normalizedValue = Math.signum(value);
-                }
-                else {
-                    normalizedValue = value/maxAbsolute;
-                }
-
-                xData.put(column, normalizedValue);
+                xData.put(column, scale(value, maxAbsolute));
                 modified = true;
             }
 
             if(scaleResponse && yData != null) {
-                Double maxAbsolute = maxAbsoluteColumnValues.get(Dataframe.COLUMN_NAME_Y);
-
                 Double value = TypeInference.toDouble(yData);
+                Double maxAbsolute = maxAbsoluteColumnValues.get(Dataframe.COLUMN_NAME_Y);
 
-                if(maxAbsolute.equals(0.0)) {
-                    yData = Math.signum(value);
-                }
-                else {
-                    yData = TypeInference.toDouble(yData)/maxAbsolute;
-                }
-
+                yData = scale(value, maxAbsolute);
                 modified = true;
             }
 
@@ -175,4 +160,19 @@ protected void _transform(Dataframe newData) {
         });
     }
 
+    /**
+     * Performs the actual rescaling handling corner cases.
+     *
+     * @param value
+     * @param maxAbsolute
+     * @return
+     */
+    private double scale(Double value, Double maxAbsolute) {
+        if(maxAbsolute.equals(0.0)) {
+            return Math.signum(value);
+        }
+        else {
+            return value/maxAbsolute;
+        }
+    }
 }
diff --git a/...src/main/java/com/datumbox/framework/core/machinelearning/preprocessing/MinMaxScaler.java b/...src/main/java/com/datumbox/framework/core/machinelearning/preprocessing/MinMaxScaler.java
@@ -167,31 +167,16 @@ protected void _transform(Dataframe newData) {
                 Double min = entry.getValue();
                 Double max = maxColumnValues.get(column);
 
-                double normalizedValue;
-                if(min.equals(max)) {
-                    normalizedValue = (value>max || (value==max && value!=0.0))?1.0:0.0;
-                }
-                else {
-                    normalizedValue = (value-min)/(max-min);
-                }
-
-                xData.put(column, normalizedValue);
+                xData.put(column, scale(value, min, max));
                 modified = true;
             }
 
             if(scaleResponse && yData != null) {
+                Double value = TypeInference.toDouble(yData);
                 Double min = minColumnValues.get(Dataframe.COLUMN_NAME_Y);
                 Double max = maxColumnValues.get(Dataframe.COLUMN_NAME_Y);
 
-                Double value = TypeInference.toDouble(yData);
-
-                if(min.equals(max)) {
-                    yData = (value>max || (value==max && value!=0.0))?1.0:0.0;
-                }
-                else {
-                    yData = (value-min)/(max-min);
-                }
-
+                yData = scale(value, min, max);
                 modified = true;
             }
 
@@ -205,4 +190,29 @@ protected void _transform(Dataframe newData) {
         });
     }
 
+    /**
+     * Performs the actual rescaling handling corner cases.
+     *
+     * @param value
+     * @param min
+     * @param max
+     * @return
+     */
+    private double scale(Double value, Double min, Double max) {
+        if(min.equals(max)) {
+            if(value>max) {
+                return 1.0;
+            }
+            else if(value==max && value!=0.0) {
+                return 1.0;
+            }
+            else {
+                return 0.0;
+            }
+        }
+        else {
+            return (value-min)/(max-min);
+        }
+    }
+
 }
diff --git a/...c/main/java/com/datumbox/framework/core/machinelearning/preprocessing/StandardScaler.java b/...c/main/java/com/datumbox/framework/core/machinelearning/preprocessing/StandardScaler.java
@@ -167,47 +167,16 @@ protected void _transform(Dataframe newData) {
                 Double mean = entry.getValue();
                 Double std = stdColumnValues.get(column);
 
-                double normalizedValue;
-                if(std.equals(0.0)) {
-                    if(value == 0.0) {
-                        normalizedValue = 0.0;
-                    }
-                    else if(value >= mean) {
-                        normalizedValue = 1.0;
-                    }
-                    else {
-                        normalizedValue = -1.0;
-                    }
-                }
-                else {
-                    normalizedValue = (value-mean)/std;
-                }
-
-                xData.put(column, normalizedValue);
+                xData.put(column, scale(value, mean, std));
                 modified = true;
             }
 
             if(scaleResponse && yData != null) {
+                Double value = TypeInference.toDouble(yData);
                 Double mean = meanColumnValues.get(Dataframe.COLUMN_NAME_Y);
                 Double std = stdColumnValues.get(Dataframe.COLUMN_NAME_Y);
 
-                Double value = TypeInference.toDouble(yData);
-
-                if(std.equals(0.0)) {
-                    if(value == 0.0) {
-                        yData = 0.0;
-                    }
-                    else if(value >= mean) {
-                        yData = 1.0;
-                    }
-                    else {
-                        yData = -1.0;
-                    }
-                }
-                else {
-                    yData = (value-mean)/std;
-                }
-
+                yData = scale(value, mean, std);
                 modified = true;
             }
 
@@ -221,4 +190,29 @@ else if(value >= mean) {
         });
     }
 
+    /**
+     * Performs the actual rescaling handling corner cases.
+     *
+     * @param value
+     * @param mean
+     * @param std
+     * @return
+     */
+    private double scale(Double value, Double mean, Double std) {
+        if(std.equals(0.0)) {
+            if(value > mean) {
+                return 1.0;
+            }
+            else if(value < mean) {
+                return -1.0;
+            }
+            else {
+                return Math.signum(value);
+            }
+        }
+        else {
+            return (value-mean)/std;
+        }
+    }
+
 }