diff --git a/stability_analysis/src/stability_analysis/core.py b/stability_analysis/src/stability_analysis/core.py
new file mode 100644
index 0000000..da02f17
--- /dev/null
+++ b/stability_analysis/src/stability_analysis/core.py
@@ -0,0 +1,93 @@
+from __future__ import annotations
+import json
+import logging
+from dataclasses import dataclass
+from typing import List, Dict, Any
+import numpy as np
+import pandas as pd
+from statistics import mean, stdev
+
+logger = logging.getLogger(__name__)
+logging.basicConfig(level=logging.INFO, format='[%(levelname)s] %(message)s')
+
+
+@dataclass
+class AnalysisResults:
+    """Strukturierte Ergebnisse der Stabilitätsanalyse."""
+    stable_cluster: List[str]
+    outlier_counts: Dict[str, int]
+    patterns: List[str]
+
+    def to_dict(self) -> Dict[str, Any]:
+        return {
+            'stable_cluster': self.stable_cluster,
+            'outlier_counts': self.outlier_counts,
+            'patterns': self.patterns,
+        }
+
+
+class InputValidationError(Exception):
+    """Wird ausgelöst, wenn Eingabedaten fehlerhaft oder unvollständig sind."""
+    pass
+
+
+def _validate_data_list(data_list: List[Dict[str, Any]]) -> None:
+    if not isinstance(data_list, list):
+        raise InputValidationError("data_list muss eine Liste von Dicts sein.")
+    if not data_list:
+        raise InputValidationError("data_list darf nicht leer sein.")
+    for i, item in enumerate(data_list):
+        if not isinstance(item, dict):
+            raise InputValidationError(f"Eintrag {i} ist kein Dict.")
+        if len(item) == 0:
+            raise InputValidationError(f"Eintrag {i} ist leer.")
+
+
+def analyze_data(data_list: List[Dict[str, Any]]) -> Dict[str, Any]:
+    """Analysiert Laufzeitdaten, um stabile Cluster, Ausreißer und Muster zu erkennen."""
+    logger.info("Starte Datenanalyse für %d Einträge.", len(data_list))
+    _validate_data_list(data_list)
+
+    try:
+        df = pd.DataFrame(data_list)
+    except Exception as e:
+        logger.error("Fehler beim Erstellen des DataFrame: %s", e)
+        raise InputValidationError("Ungültige Datenstruktur.") from e
+
+    numeric_cols = df.select_dtypes(include=[np.number]).columns.tolist()
+    if not numeric_cols:
+        raise InputValidationError("Keine numerischen Spalten für Analyse gefunden.")
+
+    stable_clusters: List[str] = []
+    outlier_counts: Dict[str, int] = {}
+    patterns: List[str] = []
+
+    for col in numeric_cols:
+        values = df[col].dropna().to_numpy()
+        if len(values) < 2:
+            continue
+        col_mean = np.mean(values)
+        col_std = np.std(values)
+        lower, upper = col_mean - 2 * col_std, col_mean + 2 * col_std
+        outlier_mask = (values < lower) | (values > upper)
+        outlier_count = int(np.sum(outlier_mask))
+        outlier_counts[col] = outlier_count
+
+        rel_std = (col_std / col_mean) if col_mean != 0 else float('inf')
+        if rel_std < 0.05:
+            stable_clusters.append(col)
+            patterns.append(f"{col}: sehr stabil (rel_std={rel_std:.3f})")
+        elif rel_std < 0.15:
+            patterns.append(f"{col}: moderat stabil (rel_std={rel_std:.3f})")
+        else:
+            patterns.append(f"{col}: variabel (rel_std={rel_std:.3f})")
+
+    result = AnalysisResults(
+        stable_cluster=stable_clusters,
+        outlier_counts=outlier_counts,
+        patterns=patterns,
+    )
+
+    logger.info("Analyse abgeschlossen. %d stabile Cluster gefunden.", len(stable_clusters))
+    assert isinstance(result.to_dict(), dict), "Analyseergebnis muss dict sein"
+    return result.to_dict()
\ No newline at end of file