Updates from Old PC

lkev · Jul 11, 2016 · 7094406 · 7094406
1 parent fbe3e70
commit 7094406
Show file tree

Hide file tree

Showing 3 changed files with 47 additions and 21 deletions.
diff --git a/fault_vs_all_example.py b/fault_vs_all_example.py
@@ -5,25 +5,25 @@
 scada = Turbine.scada_data
 
 # This gets all the data EXCEPT the faults listed. Labels as nf for "no-fault"
-nf = Turbine.filter(scada,Turbine.status_data_wec, "Main_Status",
-                     'fault_case_1', True, 600, 600, [62, 9, 80])
+nf = Turbine.filter(scada, Turbine.status_data_wec, "Main_Status",
+                    'fault_case_1', True, 600, 600, [62, 9, 80])
 # feeding fault
-ff = Turbine.filter(scada,Turbine.status_data_wec, "Main_Status",
+ff = Turbine.filter(scada, Turbine.status_data_wec, "Main_Status",
                     'fault_case_1', False, 600, 600, 62)
 # mains failure fault
-mf = Turbine.filter(scada,Turbine.status_data_wec, "Main_Status",
+mf = Turbine.filter(scada, Turbine.status_data_wec, "Main_Status",
                     'fault_case_1', False, 600, 600, 60)
 
 # generator heating fault
-gf = Turbine.filter(scada,Turbine.status_data_wec, "Main_Status",
+gf = Turbine.filter(scada, Turbine.status_data_wec, "Main_Status",
                     'fault_case_1', False, 600, 600, 9)
 
 # aircooling fault
-af = Turbine.filter(scada,Turbine.status_data_wec, "Main_Status",
+af = Turbine.filter(scada, Turbine.status_data_wec, "Main_Status",
                     'fault_case_1', False, 600, 600, 228)
 
 # excitation fault
-ef = Turbine.filter(scada,Turbine.status_data_wec, "Main_Status",
+ef = Turbine.filter(scada, Turbine.status_data_wec, "Main_Status",
                     'fault_case_1', False, 600, 600, 80)
 
 features = ['WEC_ava_windspeed',
@@ -69,7 +69,8 @@
 #                     'fault_case_1', True, 600,600,[62])
 
 # label and split into train, test and balanced training data
-xtrain, xtest, ytrain, ytest, xbaltrain, ybaltrain = Turbine.get_test_train_data(features, faults, nf)
+xtrain, xtest, ytrain, ytest, xbaltrain, ybaltrain = \
+    Turbine.get_test_train_data(features, faults, nf)
 # labels for confusion matrix
 labels = ['no-fault', 'feeding fault', 'excitation fault', 'generator fault']
 # train and test svm

diff --git a/results_fault_diagnosis_1_v_all.py b/results_fault_diagnosis_1_v_all.py
@@ -1,4 +1,6 @@
 import winfault
+import warnings
+from sklearn.grid_search import GridSearchCV, RandomizedSearchCV
 
 Turbine = winfault.WT_data()
 
@@ -49,14 +51,24 @@
 # select the faults to include.
 faults = [ff, ef, gf]
 
+# warnings suppressed because there's loads of UndefinedMetricWarning's
+warnings.filterwarnings("ignore")
+
 # label and split into train, test and balanced training data
 xtrain, xtest, ytrain, ytest, xbaltrain, ybaltrain = \
     Turbine.get_test_train_data(features, faults, nf)
+
 # labels for confusion matrix
 labels = ['no-fault', 'feeding fault', 'excitation fault', 'generator fault']
+
+# set the parameter space (class_weight is None for the balanced training data)
+parameter_space = {
+    'kernel': ['linear', 'rbf', 'poly'], 'gamma': ['auto', 1e-3, 1e-4],
+    'C': [0.01, .1, 1, 10, 100, 1000],
+    'class_weight': [{0: 0.01}, {1: 1}, {1: 2}, {1: 10}, {1: 50}, 'balanced']}
+
 # train and test svm
-print("results for basic SVM")
-winfault.svm_class_and_score(xbaltrain, ybaltrain, xtest, ytest, labels)
-print("results for bagging SVM")
-winfault.svm_class_and_score(xbaltrain, ybaltrain, xtest, ytest, labels,
-                             bagged=True)
+clf, bgg = winfault.svm_class_and_score(
+    xbaltrain, ybaltrain, xtest, ytest, labels,
+    parameter_space=parameter_space, bagged=True, score='precision_weighted',
+    search_type=GridSearchCV)
diff --git a/winfault.py b/winfault.py
@@ -888,10 +888,11 @@ def get_test_train_data(
 def svm_class_and_score(
     X_train, y_train, X_test, y_test, labels, search_type=RandomizedSearchCV,
     parameter_space={
-        'kernel': ['linear'], 'gamma': ['auto', 1e-3, 1e-4],
+        'kernel': ['linear', 'rbf', 'poly'], 'gamma': ['auto', 1e-3, 1e-4],
         'C': [0.01, .1, 1, 10, 100, 1000],
-        'class_weight': [{0: 0.01}, {1: 1}, {1: 2}, {1: 10}, {1: 50}]},
-        score='recall_weighted', iid=True, bagged=False):
+        'class_weight': [
+            {0: 0.01}, {1: 1}, {1: 2}, {1: 10}, {1: 50}, 'balanced']},
+        score='recall_weighted', iid=True, bagged=False, svm_results=True):
     """Build an SVM and return its scoring metrics
     """
     print("# Tuning hyper-parameters for %s" % score)
@@ -900,18 +901,30 @@ def svm_class_and_score(
     # Find the Hyperparameters
     clf = search_type(SVC(C=1), parameter_space, cv=10,
                       scoring=score, iid=iid)
-    if bagged is True:
-        clf = BaggingClassifier(base_estimator=clf)
 
     # Build the SVM
     clf.fit(X_train, y_train)
+    print("Hyperparameters found:")
+    print(clf.best_params_)
 
     # Make the predictions
     y_pred = clf.predict(X_test)
-
+    print()
+    print()
+    print("Results for basic SVM")
     clf_scoring(y_test, y_pred, labels)
 
-    return clf
+    if bagged is True:
+        bgg = BaggingClassifier(base_estimator=clf)
+        bgg.fit(X_train, y_train)
+        y_pred = bgg.predict(X_test)
+        print()
+        print()
+        print("Results for bagging:")
+        clf_scoring(y_test, y_pred, labels)
+        return clf, bgg
+    else:
+        return clf
 
 
 def clf_scoring(y_test, y_pred, labels):
@@ -925,7 +938,7 @@ def clf_scoring(y_test, y_pred, labels):
     cm_normalized = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]
 
     # Also print specificity metric
-    print("Specificity:", cm[0, 0] / (cm[0, 1] + cm[0, 0]))
+    # print("Specificity:", cm[0, 0] / (cm[0, 1] + cm[0, 0]))
     print(cm)
 
     # plot the confusion matrices