update markdown cells to clarify fidelity computation and K-means clustering explanation

Author: Tkemper2

results_user_community.ipynb

@@ -2644,12 +2644,13 @@
    ]
   },
   {
-   "cell_type": "markdown",
-   "id": "6c760ae2652933c4",
    "metadata": {},
+   "cell_type": "markdown",
    "source": [
-    "# TODO: show formulas for normalized entropy, fidelity, category entropy"
-   ]
+    "Remark:\n",
+    "At first we computed fidelity as 1 - normalized_entropy, but later we decided to just use normalized_entropy directly so that higher values indicate more diversity (as for category entropy). We change the naming accordingly except in some part of the code and for the naming of some files."
+   ],
+   "id": "c13e8b6121fe08dd"
   },
   {
    "cell_type": "markdown",
@@ -2846,6 +2847,12 @@
     "## 2.5 K-means clustering of groups"
    ]
   },
+  {
+   "metadata": {},
+   "cell_type": "markdown",
+   "source": "To cluster the groups based on their features -> fidelity, category entropy, and number of channels, we use the K-means clustering algorithm. This helps us identify distinct profiles of commenting behavior among the groups. Based on our analysis, K=10 clusters is a good choice allowing to get meaningful profiles without overfitting.",
+   "id": "cdfea83475361bcc"
+  },
   {
    "cell_type": "markdown",
    "id": "74ae8700d2d526e1",