Update pages/6_Feature_Engineering.py

pages/6_Feature_Engineering.py

@@ -649,4 +649,57 @@ elif file_type == "Term Frequency - Inverse Document Frequency(TF-IDF)":
     unsafe_allow_html=True,
     )
 
+    st.subheader(":red[Why log is used]")
+    st.markdown("<h1 class='title'>📌 Understanding TF-IDF Scaling</h1>", unsafe_allow_html=True)
+
+    st.markdown(
+    """
+    <div class='box'>
+        <h3 style='color: #6A0572;'>📊 Minimum and Maximum Values of N/n</h3>
+        <ul>
+            <li>When <strong>n is maximum</strong> → <span class='highlight'>N/n = 1</span></li>
+            <li>At <strong>training time</strong>: <span class='highlight'>1 ≤ n ≤ N</span></li>
+            <li>At <strong>test time</strong>: <span class='highlight'>0 ≤ n ≤ N</span> (due to Out-of-Vocabulary words)</li>
+        </ul>
+    </div>
+    """,
+    unsafe_allow_html=True,
+    )
+
+    st.markdown(
+    """
+    <div class='box'>
+        <h3 style='color: #6A0572;'>⚖️ IDF Dominance Over TF</h3>
+        <ul>
+            <li>If <strong>n decreases</strong> → <span class='highlight'>N/n increases (max)</span></li>
+            <li>TF scale is very <span class='highlight'>small</span>, but IDF scale is very <span class='highlight'>high</span></li>
+            <li>IDF can <span class='highlight'>dominate</span> TF, favoring rare words over frequent ones</li>
+        </ul>
+    </div>
+    """,
+    unsafe_allow_html=True,
+    )
+
+    st.markdown(
+    """
+    <div class='box'>
+        <h3 style='color: #6A0572;'>🛠️ How Log Solves IDF Dominance?</h3>
+        <ul>
+            <li>Applying <span class='highlight'>log</span> reduces the dominance of IDF</li>
+            <li>Logarithm <span class='highlight'>rounds off</span> values to a balanced scale</li>
+            <li>It prevents bias towards rare words and maintains proportionality</li>
+        </ul>
+    </div>
+    """,
+    unsafe_allow_html=True,
+    )
+
+    st.markdown(
+    """
+    <div class='formula'>
+        <strong>TF balances frequent words, while log(IDF) prevents rare-word dominance! 🚀</strong>
+    </div>
+    """,
+    unsafe_allow_html=True,
+    )