de:modul:m245:learningunits:lu02:loesungen:l02

Unterschiede

Hier werden die Unterschiede zwischen zwei Versionen angezeigt.

Link zu dieser Vergleichsansicht

Beide Seiten der vorigen Revision Vorhergehende Überarbeitung
Nächste Überarbeitung		 Vorhergehende Überarbeitung
de:modul:m245:learningunits:lu02:loesungen:l02 [2026/01/05 13:32] – vdemirde:modul:m245:learningunits:lu02:loesungen:l02 [2026/04/08 08:43] (aktuell) – [Modellvergleich] vdemir
Zeile 3: Zeile 3:
 ===== Voraussetzung ===== ===== Voraussetzung =====
  
-  pip install pandas scikit-learn joblib +<code python> 
-  +pip install pandas scikit-learn joblib 
 +</code> 
 ===== Python-Skript: ml_basics_shop.py ===== ===== Python-Skript: ml_basics_shop.py =====
 +<code python>
 +import pandas as pd
 +from sklearn.model_selection import train_test_split
 +from sklearn.pipeline import Pipeline
 +from sklearn.preprocessing import StandardScaler
 +from sklearn.linear_model import LogisticRegression
 +from sklearn.tree import DecisionTreeClassifier
 +from sklearn.metrics import accuracy_score, confusion_matrix, classification_report
 +import joblib
 + 
 +# -----------------------------  
 +# Daten laden
 +# -----------------------------
 +data = pd.read_csv("shop_data.csv")
 +X = data.drop("buy", axis=1)
 +y = data["buy"]
  
-  import pandas as pd +# ----------------------------- 
-  from sklearn.model_selection import train_test_split +# Train / Test Split 
-  from sklearn.pipeline import Pipeline +# ----------------------------- 
-  from sklearn.preprocessing import StandardScaler +X_train, X_test, y_train, y_test = train_test_split( X, y, test_size=0.2, random_state=42) 
-  from sklearn.linear_model import LogisticRegression + 
-  from sklearn.tree import DecisionTreeClassifier +# ----------------------------- 
-  from sklearn.metrics import accuracy_score, confusion_matrix, classification_report +# Modell 1: Logistische Regression 
-  import joblib +# ----------------------------- 
-  #  +log_reg_pipeline = Pipeline([ 
-  # -----------------------------   +    ("scaler", StandardScaler()), 
-  # Daten laden +    ("model", LogisticRegression()) 
-  # ----------------------------- +]) 
-  data = pd.read_csv("shop_data.csv") + 
-  # +log_reg_pipeline.fit(X_train, y_train) 
-  X = data.drop("buy", axis=1) +y_pred_lr = log_reg_pipeline.predict(X_test) 
-  y = data["buy"] +# 
-  # +print("Logistische Regression") 
-  # ----------------------------- +print("Accuracy:", accuracy_score(y_test, y_pred_lr)) 
-  # Train / Test Split +print("Confusion Matrix:\n", confusion_matrix(y_test, y_pred_lr)) 
-  # ----------------------------- +print("Classification Report:\n", classification_report(y_test, y_pred_lr)) 
-  X_train, X_test, y_train, y_test = train_test_split( X, y, test_size=0.2, random_state=42) + 
-  # +# ----------------------------- 
-  # ----------------------------- +# Modell 2: Decision Tree 
-  # Modell 1: Logistische Regression +# ----------------------------- 
-  # ----------------------------- +tree_model = DecisionTreeClassifier(random_state=42) 
-  log_reg_pipeline = Pipeline([ +tree_model.fit(X_train, y_train) 
-      ("scaler", StandardScaler()), +y_pred_tree = tree_model.predict(X_test) 
-      ("model", LogisticRegression()) + 
-  ]) +print("\nDecision Tree") 
-  # +print("Accuracy:", accuracy_score(y_test, y_pred_tree)) 
-  log_reg_pipeline.fit(X_train, y_train) +print("Confusion Matrix:\n", confusion_matrix(y_test, y_pred_tree)) 
-  y_pred_lr = log_reg_pipeline.predict(X_test) +print("Classification Report:\n", classification_report(y_test, y_pred_tree)) 
-  # + 
-  print("Logistische Regression") +# ----------------------------- 
-  print("Accuracy:", accuracy_score(y_test, y_pred_lr)) +# Bestes Modell speichern 
-  print("Confusion Matrix:\n", confusion_matrix(y_test, y_pred_lr)) +# ----------------------------- 
-  print("Classification Report:\n", classification_report(y_test, y_pred_lr)) +joblib.dump(log_reg_pipeline, "best_model.joblib") 
-  # + 
-  # ----------------------------- +# ----------------------------- 
-  # Modell 2: Decision Tree +# Neue Vorhersage 
-  # ----------------------------- +# ----------------------------- 
-  tree_model = DecisionTreeClassifier(random_state=42) +new_customer = pd.DataFrame([{ 
-  tree_model.fit(X_train, y_train) +    "age": 32, 
-  y_pred_tree = tree_model.predict(X_test) +    "past_purchases": 5, 
-  # +    "minutes_on_page": 6.5 
-  print("\nDecision Tree") +}]) 
-  print("Accuracy:", accuracy_score(y_test, y_pred_tree)) + 
-  print("Confusion Matrix:\n", confusion_matrix(y_test, y_pred_tree)) +loaded_model = joblib.load("best_model.joblib") 
-  print("Classification Report:\n", classification_report(y_test, y_pred_tree)) +prediction = loaded_model.predict(new_customer) 
-  # + 
-  # ----------------------------- +print("\nVorhersage fuer neuen Kunden:", prediction[0]) 
-  # Bestes Modell speichern +</code>
-  # ----------------------------- +
-  joblib.dump(log_reg_pipeline, "best_model.joblib") +
-  # +
-  # ----------------------------- +
-  # Neue Vorhersage +
-  # ----------------------------- +
-  new_customer = pd.DataFrame([{ +
-      "age": 32, +
-      "past_purchases": 5, +
-      "minutes_on_page": 6.5 +
-  }]) +
-  # +
-  loaded_model = joblib.load("best_model.joblib") +
-  prediction = loaded_model.predict(new_customer) +
-  # +
-  print("\nVorhersage fuer neuen Kunden:", prediction[0])+
      
-   +===== Modellvergleich =====
-===== 3. Modellvergleich =====+
 ^ Kriterium ^ Logistische Regression ^ Decision Tree ^ ^ Kriterium ^ Logistische Regression ^ Decision Tree ^
 | Interpretierbarkeit | hoch | mittel |  | Interpretierbarkeit | hoch | mittel | 
 | Overfitting-Gefahr | gering | hoch |  | Overfitting-Gefahr | gering | hoch | 
-| Skalierung | noetig ja | nein | +| Skalierung | nötig ja | nein | 
 | Didaktisch | sinnvoll sehr | ja |  | Didaktisch | sinnvoll sehr | ja | 
  
-Fazit: +===== Fazit ===== 
-Bei kleinen, sauberen Datensaetzen ist die Logistische Regression meist stabiler. +  * Bei kleinen, sauberen Datensätzen ist die Logistische Regression meist stabiler.  
-Decision Trees sind anschaulich, aber uebermotiviert – sie merken sich gern alles.+  * Decision Trees sind anschaulich, aber übermotiviert – sie merken sich gern alles bzw. "lernen einen Fall auswendig". 
 + 
 +---- 
 +[[https://creativecommons.org/licenses/by-nc-sa/4.0/|{{https://i.creativecommons.org/l/by-nc-sa/4.0/88x31.png}}]] Volkan Demir 
      
  • de/modul/m245/learningunits/lu02/loesungen/l02.1767616342.txt.gz
  • Zuletzt geändert: 2026/01/05 13:32
  • von vdemir