Create 16_Metrics.py

pages/16_Metrics.py

@@ -0,0 +1,122 @@
+import streamlit as st
+
+st.set_page_config(page_title="Model Evaluation Metrics", page_icon="📊", layout="wide")
+
+# Custom styling
+st.markdown("""
+    <style>
+        .stApp {
+            background-color: #1e1e1e;
+            color: white;
+        }
+        h1, h2, h3 {
+            color: #FF4C60;
+        }
+        .sidebar .sidebar-content {
+            background-color: #1e1e1e;
+        }
+        a {
+            color: #58a6ff;
+        }
+    </style>
+""", unsafe_allow_html=True)
+
+st.sidebar.title("📊 Evaluation Metrics")
+st.sidebar.markdown("Learn how to evaluate model performance in classification and regression.")
+
+# Title
+st.markdown("<h1 style='text-align: center;'>📏 Model Evaluation Metrics</h1>", unsafe_allow_html=True)
+
+# Classification Metrics
+with st.expander("🎯 Classification Metrics"):
+    st.write("""
+    Classification metrics help assess how well your model performs in classifying data correctly.
+    """)
+
+    st.markdown("### 1. Accuracy")
+    st.write("""
+    Accuracy = (Correct Predictions) / (Total Predictions)
+
+    ⚠️ Don't use accuracy if your dataset is **imbalanced** or predictions are **probabilistic**.
+    """)
+
+    st.markdown("### 2. Confusion Matrix")
+    st.write("""
+    A confusion matrix shows actual vs predicted classifications:
+
+    |       | Predicted Positive | Predicted Negative |
+    |-------|--------------------|--------------------|
+    | Actual Positive | TP (True Positive) | FN (False Negative) |
+    | Actual Negative | FP (False Positive) | TN (True Negative) |
+
+    ✅ Use this when you have **imbalanced classes**.
+    ⚠️ Don't use if your model outputs probabilities.
+    """)
+
+    st.markdown("### 3. Precision")
+    st.latex(r"Precision = \frac{TP}{TP + FP}")
+    st.write("Precision is the proportion of true positives among all predicted positives.")
+
+    st.markdown("### 4. Recall")
+    st.latex(r"Recall = \frac{TP}{TP + FN}")
+    st.write("Recall is the proportion of actual positives correctly identified.")
+
+    st.markdown("### 5. F1 Score")
+    st.latex(r"F1 = 2 \cdot \frac{Precision \cdot Recall}{Precision + Recall}")
+    st.write("F1 Score is the harmonic mean of Precision and Recall.")
+
+    st.markdown("### 6. ROC Curve & AUC")
+    st.write("""
+    - **ROC Curve**: Plot of TPR vs. FPR.
+    - **AUC**: Area Under ROC Curve → higher is better.
+
+    Ideal ROC curve hugs the top-left corner.
+    """)
+
+    st.markdown("### 7. Log Loss")
+    st.latex(r"LogLoss = -\frac{1}{n} \sum \left[ y \log(\hat{y}) + (1 - y) \log(1 - \hat{y}) \right]")
+    st.write("""
+    - Penalizes wrong predictions more if they're confident.
+    - Best for **probability-based models**.
+
+    🔥 Lower Log Loss = better performance.
+    """)
+
+# Regression Metrics
+with st.expander("📈 Regression Metrics"):
+    st.write("Evaluate how close the predictions are to the actual continuous values.")
+
+    st.markdown("### 1. Mean Squared Error (MSE)")
+    st.latex(r"MSE = \frac{1}{n} \sum_{i=1}^{n} (y_i - \hat{y}_i)^2")
+    st.write("Measures average squared difference. Sensitive to outliers.")
+
+    st.markdown("### 2. Mean Absolute Error (MAE)")
+    st.latex(r"MAE = \frac{1}{n} \sum_{i=1}^{n} |y_i - \hat{y}_i|")
+    st.write("Measures average absolute difference. More robust to outliers.")
+
+    st.markdown("### 3. Root Mean Squared Error (RMSE)")
+    st.latex(r"RMSE = \sqrt{MSE}")
+    st.write("Same as MSE, but in original units.")
+
+    st.markdown("### 4. R² Score (Coefficient of Determination)")
+    st.latex(r"R^2 = 1 - \frac{SS_{res}}{SS_{tot}}")
+    st.write("""
+    Indicates how well the model explains the variance:
+
+    - **R² = 1** → Perfect model  
+    - **0 < R² < 1** → Good model  
+    - **R² = 0** → No better than the mean  
+    - **R² < 0** → Worse than just predicting the mean  
+    """)
+
+# Summary
+st.markdown("---")
+st.markdown("### ✅ Choosing the Right Metric")
+st.write("""
+- For **Classification**: Use **F1-score**, **Log Loss**, and **Confusion Matrix**.
+- For **Regression**: Use **R²**, **MAE**, or **RMSE**.
+- ⚠️ **Avoid accuracy** in imbalanced datasets or when predicting probabilities.
+- Always compare your model against a **baseline (dummy) model**.
+""")
+
+st.success("By understanding metrics well, you can evaluate and improve your models with confidence!")