Update app.py

app.py

@@ -4,7 +4,7 @@ st.set_page_config(page_title="ML Performance Metrics", layout="wide")
 
 # Title
 st.title("📊 Machine Learning Performance Metrics")
-st.markdown("Understand key metrics used to evaluate ML models — for both **classification** and **regression** problems.")
+st.markdown("Learn how to evaluate ML models for both **classification** and **regression** problems with detailed explanations, formulas, and tips.")
 
 st.markdown("---")
 
@@ -14,60 +14,78 @@ tab1, tab2 = st.tabs(["🧠 Classification Metrics", "📈 Regression Metrics"])
 # ======================== CLASSIFICATION ========================
 with tab1:
     st.header("🧠 Classification Metrics")
-    st.markdown("These metrics evaluate how well a model classifies input data into distinct categories.")
+    st.markdown("Classification metrics help evaluate how well your model predicts categories or labels.")
+
+    with st.expander("🔍 Understanding TP, TN, FP, FN"):
+        st.markdown("""
+        - **True Positive (TP)**: Model predicted Positive and it was actually Positive  
+        - **True Negative (TN)**: Model predicted Negative and it was actually Negative  
+        - **False Positive (FP)**: Model predicted Positive but it was actually Negative (Type I Error)  
+        - **False Negative (FN)**: Model predicted Negative but it was actually Positive (Type II Error)
+        """)
+        st.image("https://upload.wikimedia.org/wikipedia/commons/thumb/2/26/Precisionrecall.svg/1200px-Precisionrecall.svg.png", caption="TP, FP, TN, FN Overview", use_column_width=True)
 
     with st.expander("🎯 Accuracy"):
         st.latex(r"Accuracy = \frac{TP + TN}{TP + TN + FP + FN}")
-        st.markdown("Proportion of correct predictions out of all predictions.")
-        st.info("✔️ Best used when classes are balanced.")
+        st.markdown("**How often the model is correct overall.**")
+        st.info("✔️ Best when classes are balanced. Can be misleading for imbalanced datasets.")
 
     with st.expander("🎯 Precision"):
         st.latex(r"Precision = \frac{TP}{TP + FP}")
-        st.markdown("How many predicted positives are actually correct?")
-        st.info("✔️ Important when False Positives are costly (e.g., spam detection).")
+        st.markdown("**Out of all predicted positives, how many were actually positive?**")
+        st.info("✔️ Use when False Positives are costly (e.g., spam detection).")
 
-    with st.expander("🎯 Recall (Sensitivity)"):
+    with st.expander("🎯 Recall (Sensitivity / True Positive Rate)"):
         st.latex(r"Recall = \frac{TP}{TP + FN}")
-        st.markdown("How many actual positives did we correctly predict?")
-        st.info("✔️ Critical when False Negatives are costly (e.g., disease detection).")
+        st.markdown("**Out of all actual positives, how many were correctly predicted?**")
+        st.info("✔️ Use when False Negatives are costly (e.g., disease diagnosis).")
+
+    with st.expander("🎯 Specificity (True Negative Rate)"):
+        st.latex(r"Specificity = \frac{TN}{TN + FP}")
+        st.markdown("**How many actual negatives were correctly predicted as negative?**")
+        st.info("✔️ Important when you also care about negatives being classified correctly.")
 
     with st.expander("🎯 F1 Score"):
         st.latex(r"F1 = 2 \cdot \frac{Precision \cdot Recall}{Precision + Recall}")
-        st.markdown("Harmonic mean of precision and recall.")
-        st.info("✔️ Good balance when you need both precision and recall.")
-
-    with st.expander("🎯 ROC-AUC Score"):
-        st.markdown("Area under the ROC curve — plots **True Positive Rate** vs **False Positive Rate**.")
-        st.image("https://upload.wikimedia.org/wikipedia/commons/6/6b/Roccurves.png", caption="ROC Curves Example",use_container_width=True)
-        st.info("✔️ Best when evaluating probabilistic classifiers.")
+        st.markdown("**Balances both Precision and Recall.**")
+        st.info("✔️ Useful when there's an uneven class distribution.")
+
+    with st.expander("🎯 ROC Curve & AUC"):
+        st.markdown("""
+        - **ROC Curve** plots the **True Positive Rate (Recall)** vs **False Positive Rate**.  
+        - **AUC** (Area Under the Curve) measures how well the model separates the classes.  
+        - AUC ranges from 0 to 1. Closer to 1 = better.
+        """)
+        st.image("https://upload.wikimedia.org/wikipedia/commons/6/6b/Roccurves.png", caption="ROC Curves Example", use_column_width=True)
+        st.info("✔️ Best for evaluating probabilistic classifiers.")
 
     st.markdown("---")
-    st.success("📌 Tip: Always use multiple metrics for a complete performance picture.")
+    st.success("📌 Tip: Use **F1 Score** or **ROC-AUC** in imbalanced classification problems!")
 
 # ======================== REGRESSION ========================
 with tab2:
     st.header("📈 Regression Metrics")
-    st.markdown("These metrics evaluate how well a model predicts continuous numerical values.")
+    st.markdown("Regression metrics help evaluate how well your model predicts **continuous numeric values**.")
 
     with st.expander("🎯 Mean Absolute Error (MAE)"):
         st.latex(r"MAE = \frac{1}{n} \sum_{i=1}^{n} |y_i - \hat{y}_i|")
-        st.markdown("Average of absolute differences between actual and predicted values.")
-        st.info("✔️ Less sensitive to outliers.")
+        st.markdown("**Average absolute difference** between predicted and actual values.")
+        st.info("✔️ Easy to understand. Less sensitive to outliers.")
 
     with st.expander("🎯 Mean Squared Error (MSE)"):
         st.latex(r"MSE = \frac{1}{n} \sum_{i=1}^{n} (y_i - \hat{y}_i)^2")
-        st.markdown("Average of squared differences. Penalizes larger errors more.")
-        st.info("✔️ Good for highlighting large deviations.")
+        st.markdown("**Average of squared differences** between predicted and actual values.")
+        st.info("✔️ Penalizes large errors more than MAE.")
 
     with st.expander("🎯 Root Mean Squared Error (RMSE)"):
         st.latex(r"RMSE = \sqrt{MSE}")
-        st.markdown("Square root of MSE. Same units as target variable.")
-        st.info("✔️ Interpretable on the same scale as the original data.")
+        st.markdown("**Square root of MSE**. Same unit as the target variable.")
+        st.info("✔️ More interpretable. Heavily penalizes large errors.")
 
     with st.expander("🎯 R² Score (Coefficient of Determination)"):
         st.latex(r"R^2 = 1 - \frac{\sum (y_i - \hat{y}_i)^2}{\sum (y_i - \bar{y})^2}")
-        st.markdown("Measures proportion of variance explained by the model.")
-        st.info("✔️ Ranges from 0 to 1. Closer to 1 means better model.")
+        st.markdown("**Proportion of variance in the target explained by the model.**")
+        st.info("✔️ Closer to 1 = better fit. Can be negative if model is worse than a mean predictor.")
 
     st.markdown("---")
-    st.success("📌 Tip: Use MAE when you care about average error, and RMSE when you want to penalize large errors.")
+    st.success("📌 Tip: Use **MAE** for average error insights, **RMSE** for large error sensitivity, and **R²** for overall model fit.")