merge

source/classifierManager/oulad/main.py

@@ -46,11 +46,12 @@ if __name__ == "__main__":
     #features = [5, 6, 7, 8, 9, 10, 15]
     features = ConstantFeature.FEATURES_ALL
     
-    #(accuracy, confusionMatrix, resultWeight) = scriptAB.predictionTask(idCourse, groupNames, whereToCut=None)
-
-  
+    #(accuracy, confusionMatrix, resultWeight) = scriptAB.predictionTask(idCourse, groupNames, whereToCut=25)
     #print("Accuracy:"+str(accuracy))
     #print("Confusion Matrix:"+str(confusionMatrix))
+    
+    #scriptAB.predictionTaskNTimes(idCourse, groupNames, whereToCut=None, ntime=2, cache=True)
+    #scriptAB.predictionTaskForAllCourses(groupNames, ConstantModel.ID_COURSES, whereToCut=25)
 
     scriptAB.predictionTaskForAllCoursesPeriod(groupNames, ConstantModel.ID_COURSES) 
     #scriptABTemporal.predictionTaskForAllCoursesPeriod(groups)

source/classifierManager/oulad/scriptAdaBoost.py

@@ -28,7 +28,6 @@ from featureManager.oulad.constant import Constant as ConstantFeature
 from featureManager.ocr.featureGroupProxyTemporal import FeatureGroupProxyTemporal
 
 from classifierManager.boosting import Boosting
-
 from classifierManager.oulad.scriptClassifier import ScriptClassifier
 
 class ScriptAdaBoost(ScriptClassifier):

source/classifierManager/oulad/scriptClassifier.py

@@ -91,9 +91,8 @@ class ScriptClassifier:
 
         ''' load feature ''' 
 
-        self.featureMultiGroup.featureGroups = []
-        self.featureMultiGroup.data = pd.DataFrame()
-        self.featureMultiGroup.addGroup(course, groupNames, self.features, whereToCut, whereToCutUnity)
+        self.featureMultiGroup.clear()
+        self.featureMultiGroup.addGroup(course, groupNames, self.features, whereToCut=whereToCut, whereToCutUnity=whereToCutUnity)
 
         train_x, train_y, test_x, test_y = self.featureMultiGroup.getTrainTestNumpyFeatures()
         test_y = np.ravel(test_y)
@@ -145,6 +144,7 @@ class ScriptClassifier:
         model = self.loadModel()
         
         print("Model loaded:"+str(type(model)))
+        print("groups:"+str(groups))
         
         model.fit(train_x, train_y)
         
@@ -157,7 +157,8 @@ class ScriptClassifier:
          
         print("--------------------------------------------------")
         print("input dimension:"+str(train_x.shape))
-        print("train[0, 500:600]="+str(train_x[0, 10:60]))
+        print("input dimension:"+str(train_y.shape))
+        print("train[0, 10:60]="+str(train_x[0, 10:60]))
         courseName = course.getName()
         print(("Course "+courseName+": "+str(accuracy)))
         print("Confusion Matrix:")
@@ -179,12 +180,13 @@ class ScriptClassifier:
         '''
         accuracys = []
         
+        course = Course(idCourse)
         size = len(groupNames)
         confusionMatrice = np.zeros((size, size))
         
         dictResultWeight = {}
         
-        nameCourse = Course.getCourseName(idCourse, summary=True)
+        nameCourse = course.getName()
         fileNameSavedResult = os.path.join(ScriptClassifier.DIRECTORY_EXPERIMENT, 
                                            self.fileName+
                                            str(nameCourse)+'Session'+str(whereToCut)+'.p')
@@ -244,17 +246,25 @@ class ScriptClassifier:
         
         with open(fileNameResult, 'w') as fileResult:
             for idCourse in idCourses: #[Course.COURSE_ID_ARDUINO, Course.COURSE_ID_AUDACE]
+                
+                print("idCourse: "+str(idCourse))
+                course = Course(idCourse)
+                
+
                 (accuracy, confusionMatrix, dictWeight) = self.predictionTaskNTimes(
                                                                    idCourse, 
                                                                    groupNames,
                                                                    whereToCut=whereToCut, 
-                                                                   ntime=10)
+                                                                   ntime=2)
                 
                 scores.append(accuracy)
                 
                 if dictWeight:
-                    dictWeightAllCourses[idCourse] = dictWeight    
-                courseName = Course.getCourseName(idCourse, summary=True)
+
+                    dictWeightAllCourses[idCourse] = dictWeight
+                       
+                courseName = course.getName()
+
                 fileResult.write("--------------------------\n")
                 fileResult.write("Course "+courseName+": accuracy("+str(accuracy)+") \n")
                 

source/featureManager/oulad/featureMultiGroup.py

@@ -60,7 +60,7 @@ class FeatureMultiGroup:
         
         if self.course is None:
             self.course = featureGroup.getCourse()
-        elif self.course != featureGroup.getCourse():
+        elif self.course.getIdCourse() != featureGroup.getCourse().getIdCourse():
             warnings.warn('featureManager.features - FeatureMultiGroup - FeatureMultiGroup are merged from different courses', Warning)
         label = len(self.featureGroups)
         self.featureGroups.append(featureGroup)
@@ -154,6 +154,8 @@ class FeatureMultiGroup:
     
         trainX, trainY, testX, testY = self.getTrainTestPandasFeatures(train, test)
         
-        return trainX.values, trainY.values, testX.values, testY.values
+        return trainX.values.astype(float), trainY.values.astype(int), testX.values.astype(float), testY.values.astype(int)
         
+    def clear(self):
+        self.__init__()
         
\ No newline at end of file

source/featureManager/oulad/featureSequence.py

@@ -257,7 +257,7 @@ class FeatureSequence:
         @return: a dataframe with resources as column and one row per user with the averages
         '''
         
-        return FeatureSequence.__getNumberAccessPerResource(idUser, events, course, np.mean, filterType)
+        return FeatureSequence.__getNumberAccessPerResource(idUser, events, course, lambda l : np.sum(l)/len(l) if len(l) else 0, filterType)
 
     @staticmethod
     def getAllMarks(idUser, events, course):

source/featureManager/oulad/testUnitFeatureSequence.py

@@ -43,7 +43,7 @@ class Test(unittest.TestCase):
         
         print(nameFeature+": "+str(featureWindowDuration[nameFeature].values[0]))
         
-        self.assertTrue(featureWindowDuration[nameFeature].values[0] == -259)
+        self.assertTrue(featureWindowDuration[nameFeature].values[0] == 258)
         
         dataEmpty = FeatureSequence.getWindowDuration(idUser, pd.DataFrame())
         

source/model/oulad/constant.py

@@ -78,6 +78,8 @@ class Constant:
    
     ID_COURSES = [AAA_2013_J]
                   #DDD_2013_B,DDD_2013_J,DDD_2014_B,DDD_2014_J,FFF_2014_J,FFF_2013_B,FFF_2013_J,FFF_2014_B]
+
+
       
     
     REGISTRATION = 'registration'