System save at 22/09/2025 17:03 by user_client2024

.ipynb_checkpoints/main-checkpoint.ipynb

@@ -11,6 +11,49 @@
     "import pandas as pd\n",
     "from tms_data_interface import SQLQueryInterface\n",
     "\n",
+    "def apply_sar_flag(df, var1, var2, var3, random_state=42):\n",
+    "    \"\"\"\n",
+    "    Apply percentile-based thresholds, split data into alerting and non-alerting,\n",
+    "    flag random 10% of alerting data as 'Y', and merge back.\n",
+    "\n",
+    "    Parameters:\n",
+    "        df (pd.DataFrame): Input dataframe\n",
+    "        var1 (str): First variable (for 50th percentile threshold)\n",
+    "        var2 (str): Second variable (for 50th percentile threshold)\n",
+    "        var3 (str): Third variable (for 90th percentile threshold)\n",
+    "        random_state (int): Seed for reproducibility\n",
+    "\n",
+    "    Returns:\n",
+    "        pd.DataFrame: DataFrame with 'SAR_Flag' column added\n",
+    "    \"\"\"\n",
+    "\n",
+    "    # Calculate thresholds\n",
+    "    th1 = np.percentile(df[var1].dropna(), 50)\n",
+    "    th2 = np.percentile(df[var2].dropna(), 50)\n",
+    "    th3 = np.percentile(df[var3].dropna(), 90)\n",
+    "\n",
+    "    # Split into alerting and non-alerting\n",
+    "    alerting = df[(df[var1] >= th1) &\n",
+    "                  (df[var2] >= th2) &\n",
+    "                  (df[var3] >= th3)].copy()\n",
+    "\n",
+    "    non_alerting = df.loc[~df.index.isin(alerting.index)].copy()\n",
+    "\n",
+    "    # Assign SAR_Flag = 'N' for non-alerting\n",
+    "    non_alerting['SAR_Flag'] = 'N'\n",
+    "\n",
+    "    # Assign SAR_Flag for alerting data\n",
+    "    alerting['SAR_Flag'] = 'N'\n",
+    "    n_y = int(len(alerting) * 0.1)   # 10% count\n",
+    "    if n_y > 0:\n",
+    "        y_indices = alerting.sample(n=n_y, random_state=random_state).index\n",
+    "        alerting.loc[y_indices, 'SAR_Flag'] = 'Y'\n",
+    "\n",
+    "    # Merge back and preserve original order\n",
+    "    final_df = pd.concat([alerting, non_alerting]).sort_index()\n",
+    "\n",
+    "    return final_df\n",
+    "\n",
     "query = \"\"\"\n",
     "WITH time_windows AS (\n",
     "    SELECT\n",
@@ -107,9 +150,14 @@
     "        final_scenario_df['PARTICIPANT_VOLUME_PCT'] = final_scenario_df['PARTICIPANT_VOLUME']/\\\n",
     "                                                        final_scenario_df['TOTAL_VOLUME'] * 100\n",
     "        final_scenario_df['Segment'] = 'Default'\n",
-    "        final_scenario_df['SAR_FLAG'] = 'N'\n",
+    "        # final_scenario_df['SAR_FLAG'] = 'N'\n",
     "        final_scenario_df['Risk'] = 'Medium Risk'\n",
     "        final_scenario_df.dropna(inplace=True)\n",
+    "        final_scenario_df = apply_sar_flag(final_scenario_df,\n",
+    "                                           'PRICE_CHANGE_PCT',\n",
+    "                                           'PARTICIPANT_VOLUME_PCT',\n",
+    "                                           'TOTAL_VOLUME',\n",
+    "                                           random_state=42)\n",
     "        # final_scenario_df['RUN_DATE'] = final_scenario_df['END_DATE']\n",
     "        return final_scenario_df\n"
    ]

main.ipynb

@@ -11,6 +11,49 @@
     "import pandas as pd\n",
     "from tms_data_interface import SQLQueryInterface\n",
     "\n",
+    "def apply_sar_flag(df, var1, var2, var3, random_state=42):\n",
+    "    \"\"\"\n",
+    "    Apply percentile-based thresholds, split data into alerting and non-alerting,\n",
+    "    flag random 10% of alerting data as 'Y', and merge back.\n",
+    "\n",
+    "    Parameters:\n",
+    "        df (pd.DataFrame): Input dataframe\n",
+    "        var1 (str): First variable (for 50th percentile threshold)\n",
+    "        var2 (str): Second variable (for 50th percentile threshold)\n",
+    "        var3 (str): Third variable (for 90th percentile threshold)\n",
+    "        random_state (int): Seed for reproducibility\n",
+    "\n",
+    "    Returns:\n",
+    "        pd.DataFrame: DataFrame with 'SAR_Flag' column added\n",
+    "    \"\"\"\n",
+    "\n",
+    "    # Calculate thresholds\n",
+    "    th1 = np.percentile(df[var1].dropna(), 50)\n",
+    "    th2 = np.percentile(df[var2].dropna(), 50)\n",
+    "    th3 = np.percentile(df[var3].dropna(), 90)\n",
+    "\n",
+    "    # Split into alerting and non-alerting\n",
+    "    alerting = df[(df[var1] >= th1) &\n",
+    "                  (df[var2] >= th2) &\n",
+    "                  (df[var3] >= th3)].copy()\n",
+    "\n",
+    "    non_alerting = df.loc[~df.index.isin(alerting.index)].copy()\n",
+    "\n",
+    "    # Assign SAR_Flag = 'N' for non-alerting\n",
+    "    non_alerting['SAR_Flag'] = 'N'\n",
+    "\n",
+    "    # Assign SAR_Flag for alerting data\n",
+    "    alerting['SAR_Flag'] = 'N'\n",
+    "    n_y = int(len(alerting) * 0.1)   # 10% count\n",
+    "    if n_y > 0:\n",
+    "        y_indices = alerting.sample(n=n_y, random_state=random_state).index\n",
+    "        alerting.loc[y_indices, 'SAR_Flag'] = 'Y'\n",
+    "\n",
+    "    # Merge back and preserve original order\n",
+    "    final_df = pd.concat([alerting, non_alerting]).sort_index()\n",
+    "\n",
+    "    return final_df\n",
+    "\n",
     "query = \"\"\"\n",
     "WITH time_windows AS (\n",
     "    SELECT\n",
@@ -107,9 +150,14 @@
     "        final_scenario_df['PARTICIPANT_VOLUME_PCT'] = final_scenario_df['PARTICIPANT_VOLUME']/\\\n",
     "                                                        final_scenario_df['TOTAL_VOLUME'] * 100\n",
     "        final_scenario_df['Segment'] = 'Default'\n",
-    "        final_scenario_df['SAR_FLAG'] = 'N'\n",
+    "        # final_scenario_df['SAR_FLAG'] = 'N'\n",
     "        final_scenario_df['Risk'] = 'Medium Risk'\n",
     "        final_scenario_df.dropna(inplace=True)\n",
+    "        final_scenario_df = apply_sar_flag(final_scenario_df,\n",
+    "                                           'PRICE_CHANGE_PCT',\n",
+    "                                           'PARTICIPANT_VOLUME_PCT',\n",
+    "                                           'TOTAL_VOLUME',\n",
+    "                                           random_state=42)\n",
     "        # final_scenario_df['RUN_DATE'] = final_scenario_df['END_DATE']\n",
     "        return final_scenario_df\n"
    ]

main.py

@@ -8,6 +8,49 @@ from datetime import datetime, timedelta
 import pandas as pd
 from tms_data_interface import SQLQueryInterface
 
+def apply_sar_flag(df, var1, var2, var3, random_state=42):
+    """
+    Apply percentile-based thresholds, split data into alerting and non-alerting,
+    flag random 10% of alerting data as 'Y', and merge back.
+
+    Parameters:
+        df (pd.DataFrame): Input dataframe
+        var1 (str): First variable (for 50th percentile threshold)
+        var2 (str): Second variable (for 50th percentile threshold)
+        var3 (str): Third variable (for 90th percentile threshold)
+        random_state (int): Seed for reproducibility
+
+    Returns:
+        pd.DataFrame: DataFrame with 'SAR_Flag' column added
+    """
+
+    # Calculate thresholds
+    th1 = np.percentile(df[var1].dropna(), 50)
+    th2 = np.percentile(df[var2].dropna(), 50)
+    th3 = np.percentile(df[var3].dropna(), 90)
+
+    # Split into alerting and non-alerting
+    alerting = df[(df[var1] >= th1) &
+                  (df[var2] >= th2) &
+                  (df[var3] >= th3)].copy()
+
+    non_alerting = df.loc[~df.index.isin(alerting.index)].copy()
+
+    # Assign SAR_Flag = 'N' for non-alerting
+    non_alerting['SAR_Flag'] = 'N'
+
+    # Assign SAR_Flag for alerting data
+    alerting['SAR_Flag'] = 'N'
+    n_y = int(len(alerting) * 0.1)   # 10% count
+    if n_y > 0:
+        y_indices = alerting.sample(n=n_y, random_state=random_state).index
+        alerting.loc[y_indices, 'SAR_Flag'] = 'Y'
+
+    # Merge back and preserve original order
+    final_df = pd.concat([alerting, non_alerting]).sort_index()
+
+    return final_df
+
 query = """
 WITH time_windows AS (
     SELECT
@@ -104,9 +147,14 @@ class Scenario:
         final_scenario_df['PARTICIPANT_VOLUME_PCT'] = final_scenario_df['PARTICIPANT_VOLUME']/\
                                                         final_scenario_df['TOTAL_VOLUME'] * 100
         final_scenario_df['Segment'] = 'Default'
-        final_scenario_df['SAR_FLAG'] = 'N'
+        # final_scenario_df['SAR_FLAG'] = 'N'
         final_scenario_df['Risk'] = 'Medium Risk'
         final_scenario_df.dropna(inplace=True)
+        final_scenario_df = apply_sar_flag(final_scenario_df,
+                                           'PRICE_CHANGE_PCT',
+                                           'PARTICIPANT_VOLUME_PCT',
+                                           'TOTAL_VOLUME',
+                                           random_state=42)
         # final_scenario_df['RUN_DATE'] = final_scenario_df['END_DATE']
         return final_scenario_df