Merge branch 'main' of https://gitlab.informatik.uni-halle.de/apumu/data-mining

docs/Struktur.md

@@ -111,4 +111,27 @@ VUS_PR : 0.39033909329157723
 **Evaluieren der Inferential Reproducibility**
 
 
-**Ggf. gleiches Prozedere wie bei multivariaten Daten**
\ No newline at end of file
+**Ggf. gleiches Prozedere wie bei multivariaten Daten**
+
+
+# Laufzeiten
++ Spanne gemessen zwischen POLY (voraussichtlich schnellster Algorithmus lt. Autoren) vs. KNN (voraussichtlich längster Algorithmus im Training lt. Autoren)
++ Geschätzte für alles anhand einer 101 kB Datei für Datensatz mit 321 MB
++ Getestet am:
+	+ Laptop: iCore5-8th, 4 Kerne, 1,80 GHz
+	+ Uni-Rechner: iCore5-8th, 6 Kerne, 3,40 GHz
+**KNN**
++ Einzeldatei (100kB): 
+	+ am Laptop : 265,55 s 
+	+ am Uni-Rechner: : 118.31 s
++ Laufzeit insgesamt geschätzt:
+	+ am Laptop: 852435,76 s $\approx$ 14.207 min $\approx$ 237 h $\approx$ 9,9 d
+	+ am Uni-Rechner: 379760,36 s $\approx$ 6.329,34 min $\approx$ 105,5 h $\approx$ 4,4 d
+
+**POLY**
++ Einzeldatei (100kB):
+	+ am Laptop: 180,77 s
+	+ am Uni-Rechner: 84,54 s
++ Laufzeit insgesamt geschätzt:
+	+ am Laptop: 580256,86 s $\approx$ 9.670,26 min $\approx$ 161,18 h $\approx$ 6,72 d
+	+ am Uni-Rechner: 271.356,89 s $\approx$ 4.522,62 min $\approx$ 75,38 h $\approx$ 3,14 d
\ No newline at end of file

src/get_runtime.py

+#Measure Runtime for KNN and POLY for singe File
+import sys
+import os
+import time
+BASE_DIR = os.path.abspath(os.path.join(os.path.dirname(__file__), '../'))
+if BASE_DIR not in sys.path:
+    sys.path.append(BASE_DIR)
+from src.models.sofia_modelle.KNN import KNN
+from src.models.desi.POLY import POLY
+from src.utils.slidingWindows import find_length_rank
+from src.run_model_wrapper import main
+
+#optimal hyperparameters from autors: 'POLY': {'periodicity': 1, 'power': 4}
+params = {
+        'n_neighbors': [10, 20, 30, 40, 50],
+        'method': ['largest', 'mean', 'median']
+    }
+
+def run_Sub_KNN(data, n_neighbors=10, method='largest', periodicity=1, n_jobs=1):
+    slidingWindow = find_length_rank(data, rank=periodicity)
+    clf = KNN(slidingWindow=slidingWindow, n_neighbors=n_neighbors,method=method, n_jobs=n_jobs)
+    clf.fit(data)
+    score = clf.decision_scores_
+    return score.ravel()
+
+model = 'Sub_KNN_runtime'
+output_path = '../docs/evaluation/'
+
+print('Start processing with KNN')
+start_time = time.time()
+#writes results in .csv
+main(run_Sub_KNN,params,model,data_folders = './data/', model_type='unsupervised',output_dir = output_path)
+end_time = time.time()
+duration = end_time-start_time
+print(f'Runtime for KNN: singe File {duration} s, Estimation for total dataset: {duration * 3210} s')
+
+
+
+#optimal hyperparameters from autors: 'POLY': {'periodicity': 1, 'power': 4}
+params = {
+        'periodicity': [1, 2, 3],
+        'power': [1, 2, 3, 4,5]
+}
+
+def run_POLY(data, periodicity=1, power=3, n_jobs=1):
+    slidingWindow = find_length_rank(data, rank=periodicity)
+    clf = POLY(power=power, window = slidingWindow)
+    clf.fit(data)
+    score = clf.decision_scores_
+    return score.ravel()
+
+model = 'POLY_runtime'
+
+print('Start processing with POLY')
+start_time = time.time()
+#writes results in .csv
+main(run_POLY,params,model,data_folders = './data/', model_type='unsupervised',output_dir = output_path)
+end_time = time.time()
+duration = end_time-start_time
+print(f'Runtime for POLY: singe File {duration} s, Estimation for total dataset: {duration * 3210} s')
\ No newline at end of file