SNAPSHOT 1.1 (DDM)

Vasileios Vittis · Vasileios Vittis · commit 52d042ca889c · 2021-07-21T10:32:36.000+03:00
diff --git a/README.md b/README.md
@@ -31,7 +31,11 @@ int weighted_voting_option = Integer.parseInt(params.get("weighted_voting_option
 int age_of_maturity = Integer.parseInt(params.get("age_of_maturity"));
 ```
 ``` java
-/* Drift Detection Method used */
+/* Drift Detection Method used 
+*  If drift_detection_method_id = 0 then No Concept Drift
+*  If drift_detection_method_id = 1 then DDM
+*  If drift_detection_method_id = 2 then EDDM
+*/
 int drift_detection_method_id = Integer.parseInt(params.get("drift_detection_method_id"));
 ```
 #### Kafka Input Stream
diff --git a/analysis.md b/analysis.md
@@ -79,5 +79,31 @@ Not done yet
 
 2) Implement DDM inside State
 
+I did it and uploaded the chart comparisons in Github.
+They seem all good. Except the fact that my version (Hoeffding Tree) scores better both with and without a concept drift detector. 
 
+There are two explenations for that. Either my base learner is just better, or the datasets are different.
 
+The latter is a little of because both datasets in papaer has been produced by the same generator.
+
+Let's assume that CHVT is better that C4.5
+
+
+How i did the testing. I implemented a mechanism in which when the user assigns to the variable drift_detection_method_id the value of 0 then they dictate that they do not want to use any dirft detection algorithm.
+
+Therefore, I changed the code of my pycharm script and I added two kafkaconnector listenning to two different kafka topics.
+
+I ran DistributedLearningJava without a drift detection method (by assigning 0) writing to kafka topic and then I reran my project with (option 1  aka DDM) writing it to a different kafka topic. The script reads from both of them and merges the plots into one.
+
+21/7
+
+
+'application_1614183653371_0143'.
+
+Latest Apache Flink Job http://clu02.softnet.tuc.gr:33418
+
+cd /usr/local/flink;
+
+./bin/flink run -d -p 3 -m yarn-cluster -yid application_1614183653371_0143 /home/vvittis/DistributedLearningJava/target/DistributedLearningJava-1.1-SNAPSHOT.jar --number_of_HT 1 --age_of_maturity 1000 --combination_function 3 --weighted_voting_parameter 1 --drift_detection_method_id 3
+
+already a kafka topic vvittis_SineTopic
diff --git a/dependency-reduced-pom.xml b/dependency-reduced-pom.xml
@@ -3,7 +3,7 @@
   <modelVersion>4.0.0</modelVersion>
   <groupId>org.example</groupId>
   <artifactId>DistributedLearningJava</artifactId>
-  <version>1.0-SNAPSHOT</version>
+  <version>1.1-SNAPSHOT</version>
   <build>
     <plugins>
       <plugin>
diff --git a/pom.xml b/pom.xml
@@ -6,7 +6,7 @@
 
     <groupId>org.example</groupId>
     <artifactId>DistributedLearningJava</artifactId>
-    <version>1.0-SNAPSHOT</version>
+    <version>1.1-SNAPSHOT</version>
 
     <properties>
         <project.build.sourceEncoding>UTF-8</project.build.sourceEncoding>
diff --git a/src/main/java/ConceptDriftDetector/EDDM.java b/src/main/java/ConceptDriftDetector/EDDM.java
@@ -16,4 +16,19 @@ public void FindConceptDrift(double error_rate) {
     public int getCurrentDriftStatus() {
         return 0;
     }
+
+    @Override
+    public int getSignal() {
+        return 0;
+    }
+
+    @Override
+    public void updateCurrentDriftStatus() {
+
+    }
+
+    @Override
+    public void ResetConceptDrift() {
+
+    }
 }
diff --git a/src/main/java/HoeffdingTree/HoeffdingTree.java b/src/main/java/HoeffdingTree/HoeffdingTree.java
@@ -10,6 +10,7 @@ public class HoeffdingTree implements Serializable {
     public double instances_seen;
     public double correctly_classified;
     public double weight;
+    public int tree_phase;
     public int combination_function;
     public int hoeffding_tree_id;
     public int[] m_features; // list of labels corresponding to samples of node
@@ -28,14 +29,15 @@ public HoeffdingTree() {
      *                                <p> Create the Hoeffding tree for given parameters </p>
      */
 
-    public void CreateHoeffdingTree(int m_features, int Max, int max_examples_seen, double delta, double tie_threshold, int combination_function_id, int hoeffding_tree_id) {
+    public void CreateHoeffdingTree(int m_features, int Max, int max_examples_seen, double delta, double tie_threshold, int combination_function_id, int hoeffding_tree_id, int background_tree_indicator) {
         root.CreateHT(m_features, max_examples_seen, delta, tie_threshold);
         instances_seen = 0.0;
         correctly_classified = 0.0;
         weight = 0.0;
         this.hoeffding_tree_id = hoeffding_tree_id;
         combination_function = combination_function_id;
         initialize_m_features(m_features, Max - 1, hoeffding_tree_id);
+        this.tree_phase = background_tree_indicator;
     }
 
     /**
