{"nbformat":4,"nbformat_minor":0,"metadata":{"colab":{"provenance":[],"authorship_tag":"ABX9TyOsdXEJWDJ9Hxdk7/a3voru"},"kernelspec":{"name":"python3","display_name":"Python 3"},"language_info":{"name":"python"}},"cells":[{"cell_type":"code","execution_count":12,"metadata":{"id":"62826a3f-034b-4c98-968b-7a012021d1df","executionInfo":{"status":"ok","timestamp":1766863210478,"user_tz":0,"elapsed":12,"user":{"displayName":"Salma Ahmed","userId":"10865891360196912478"}}},"outputs":[],"source":["#Section 1\n","\n","import pandas as pd\n","from sklearn.model_selection import train_test_split\n","from sklearn.tree import DecisionTreeRegressor\n","from sklearn.metrics import mean_absolute_error\n","from sklearn.metrics import mean_squared_error\n","import matplotlib.pyplot as plt\n","import numpy as np\n"]},{"cell_type":"code","source":["data = pd.read_csv(\"f1_pitstops_2018_2024.csv\")\n","data.head() # we do this to read the data\n"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":637},"id":"7fvYaR60oMfk","executionInfo":{"status":"ok","timestamp":1766863210885,"user_tz":0,"elapsed":408,"user":{"displayName":"Salma Ahmed","userId":"10865891360196912478"}},"outputId":"66a2bef4-0690-4821-8fc1-72853ea6c746"},"execution_count":13,"outputs":[{"output_type":"execute_result","data":{"text/plain":[" Season Round Circuit Driver \\\n","0 2018 1 Albert Park Grand Prix Circuit Sebastian Vettel \n","1 2018 1 Albert Park Grand Prix Circuit Sebastian Vettel \n","2 2018 1 Albert Park Grand Prix Circuit Lewis Hamilton \n","3 2018 1 Albert Park Grand Prix Circuit Lewis Hamilton \n","4 2018 1 Albert Park Grand Prix Circuit Kimi Räikkönen \n","\n"," Constructor Laps Position TotalPitStops AvgPitStopTime \\\n","0 Ferrari 58 1 1 21.787 \n","1 Ferrari 58 1 1 21.787 \n","2 Mercedes 58 2 1 21.821 \n","3 Mercedes 58 2 1 21.821 \n","4 Ferrari 58 3 1 21.421 \n","\n"," Race Name ... Tire Usage Aggression Fast Lap Attempts \\\n","0 Australian Grand Prix ... 0.017241 44.76882 \n","1 Australian Grand Prix ... 0.017241 44.76882 \n","2 Australian Grand Prix ... 0.017241 44.73482 \n","3 Australian Grand Prix ... 0.017241 44.73482 \n","4 Australian Grand Prix ... 0.017241 45.13482 \n","\n"," Position Changes Driver Aggression Score Abbreviation Stint \\\n","0 0.000000 6.755003 VET 1.0 \n","1 0.000000 6.755003 VET 2.0 \n","2 0.043478 6.754254 HAM 1.0 \n","3 0.043478 6.754254 HAM 2.0 \n","4 0.086957 6.818562 RAI 1.0 \n","\n"," Tire Compound Stint Length Pit_Lap Pit_Time \n","0 ULTRASOFT 25.0 26.0 21.787 \n","1 SOFT 32.0 NaN Final Stint \n","2 ULTRASOFT 17.0 19.0 21.821 \n","3 SOFT 39.0 NaN Final Stint \n","4 ULTRASOFT 17.0 18.0 21.421 \n","\n","[5 rows x 30 columns]"],"text/html":["\n","
\n","
\n","\n","\n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n"," \n","
SeasonRoundCircuitDriverConstructorLapsPositionTotalPitStopsAvgPitStopTimeRace Name...Tire Usage AggressionFast Lap AttemptsPosition ChangesDriver Aggression ScoreAbbreviationStintTire CompoundStint LengthPit_LapPit_Time
020181Albert Park Grand Prix CircuitSebastian VettelFerrari581121.787Australian Grand Prix...0.01724144.768820.0000006.755003VET1.0ULTRASOFT25.026.021.787
120181Albert Park Grand Prix CircuitSebastian VettelFerrari581121.787Australian Grand Prix...0.01724144.768820.0000006.755003VET2.0SOFT32.0NaNFinal Stint
220181Albert Park Grand Prix CircuitLewis HamiltonMercedes582121.821Australian Grand Prix...0.01724144.734820.0434786.754254HAM1.0ULTRASOFT17.019.021.821
320181Albert Park Grand Prix CircuitLewis HamiltonMercedes582121.821Australian Grand Prix...0.01724144.734820.0434786.754254HAM2.0SOFT39.0NaNFinal Stint
420181Albert Park Grand Prix CircuitKimi RäikkönenFerrari583121.421Australian Grand Prix...0.01724145.134820.0869576.818562RAI1.0ULTRASOFT17.018.021.421
\n","

5 rows × 30 columns

\n","
\n","
\n","\n","
\n"," \n","\n"," \n","\n"," \n","
\n","\n","\n","
\n"," \n","\n","\n","\n"," \n","
\n","\n","
\n","
\n"],"application/vnd.google.colaboratory.intrinsic+json":{"type":"dataframe","variable_name":"data"}},"metadata":{},"execution_count":13}]},{"cell_type":"code","execution_count":14,"metadata":{"id":"47d0afea-3090-4c1e-a295-54cebff972d5","executionInfo":{"status":"ok","timestamp":1766863210890,"user_tz":0,"elapsed":8,"user":{"displayName":"Salma Ahmed","userId":"10865891360196912478"}},"colab":{"base_uri":"https://localhost:8080/"},"outputId":"4a5ec710-d3bb-462a-b232-f2986f23dd89"},"outputs":[{"output_type":"stream","name":"stdout","text":["\n","RangeIndex: 7374 entries, 0 to 7373\n","Data columns (total 30 columns):\n"," # Column Non-Null Count Dtype \n","--- ------ -------------- ----- \n"," 0 Season 7374 non-null int64 \n"," 1 Round 7374 non-null int64 \n"," 2 Circuit 7374 non-null object \n"," 3 Driver 7374 non-null object \n"," 4 Constructor 7374 non-null object \n"," 5 Laps 7374 non-null int64 \n"," 6 Position 7374 non-null int64 \n"," 7 TotalPitStops 7374 non-null int64 \n"," 8 AvgPitStopTime 7189 non-null float64\n"," 9 Race Name 7001 non-null object \n"," 10 Date 7001 non-null object \n"," 11 Time_of_race 7001 non-null object \n"," 12 Location 7001 non-null object \n"," 13 Country 7001 non-null object \n"," 14 Air_Temp_C 7001 non-null float64\n"," 15 Track_Temp_C 7001 non-null float64\n"," 16 Humidity_% 7001 non-null float64\n"," 17 Wind_Speed_KMH 7001 non-null float64\n"," 18 Lap Time Variation 7189 non-null float64\n"," 19 Total Pit Stops 7374 non-null float64\n"," 20 Tire Usage Aggression 7308 non-null float64\n"," 21 Fast Lap Attempts 7189 non-null float64\n"," 22 Position Changes 7374 non-null float64\n"," 23 Driver Aggression Score 7189 non-null float64\n"," 24 Abbreviation 7374 non-null object \n"," 25 Stint 7265 non-null float64\n"," 26 Tire Compound 7265 non-null object \n"," 27 Stint Length 7265 non-null float64\n"," 28 Pit_Lap 4564 non-null float64\n"," 29 Pit_Time 7093 non-null object \n","dtypes: float64(14), int64(5), object(11)\n","memory usage: 1.7+ MB\n"]}],"source":["data.info() #we do this to see if we have any null values\n","data.dropna(inplace=True) #returns data with only non-null values, (drops rows with missing stuff)"]},{"cell_type":"code","execution_count":15,"metadata":{"id":"0064d506-6b52-4e15-9014-ee2e77a84853","executionInfo":{"status":"ok","timestamp":1766863210895,"user_tz":0,"elapsed":6,"user":{"displayName":"Salma Ahmed","userId":"10865891360196912478"}},"colab":{"base_uri":"https://localhost:8080/"},"outputId":"5fc3555c-9f71-41f5-c865-297e6de47898"},"outputs":[{"output_type":"stream","name":"stdout","text":["\n","Index: 4334 entries, 0 to 7371\n","Data columns (total 30 columns):\n"," # Column Non-Null Count Dtype \n","--- ------ -------------- ----- \n"," 0 Season 4334 non-null int64 \n"," 1 Round 4334 non-null int64 \n"," 2 Circuit 4334 non-null object \n"," 3 Driver 4334 non-null object \n"," 4 Constructor 4334 non-null object \n"," 5 Laps 4334 non-null int64 \n"," 6 Position 4334 non-null int64 \n"," 7 TotalPitStops 4334 non-null int64 \n"," 8 AvgPitStopTime 4334 non-null float64\n"," 9 Race Name 4334 non-null object \n"," 10 Date 4334 non-null object \n"," 11 Time_of_race 4334 non-null object \n"," 12 Location 4334 non-null object \n"," 13 Country 4334 non-null object \n"," 14 Air_Temp_C 4334 non-null float64\n"," 15 Track_Temp_C 4334 non-null float64\n"," 16 Humidity_% 4334 non-null float64\n"," 17 Wind_Speed_KMH 4334 non-null float64\n"," 18 Lap Time Variation 4334 non-null float64\n"," 19 Total Pit Stops 4334 non-null float64\n"," 20 Tire Usage Aggression 4334 non-null float64\n"," 21 Fast Lap Attempts 4334 non-null float64\n"," 22 Position Changes 4334 non-null float64\n"," 23 Driver Aggression Score 4334 non-null float64\n"," 24 Abbreviation 4334 non-null object \n"," 25 Stint 4334 non-null float64\n"," 26 Tire Compound 4334 non-null object \n"," 27 Stint Length 4334 non-null float64\n"," 28 Pit_Lap 4334 non-null float64\n"," 29 Pit_Time 4334 non-null object \n","dtypes: float64(14), int64(5), object(11)\n","memory usage: 1.0+ MB\n"]}],"source":["data.info() #reload the new data"]},{"cell_type":"code","execution_count":16,"metadata":{"id":"4adbc4c6-eaad-4077-b102-056315859576","executionInfo":{"status":"ok","timestamp":1766863210927,"user_tz":0,"elapsed":31,"user":{"displayName":"Salma Ahmed","userId":"10865891360196912478"}}},"outputs":[],"source":["TARGET_COLUMN = 'Lap Time Variation' # the thing we want to predict (change name if needed)\n","CONSTRUCTOR_COL = 'Constructor' # column that holds the team name\n","DATA_PATH = \"f1_pitstops_2018_2024.csv\"\n"," # path to the CSV file"]},{"cell_type":"code","execution_count":17,"metadata":{"id":"4ef19360-6855-4e3f-85e9-f41ed85df19d","executionInfo":{"status":"ok","timestamp":1766863210928,"user_tz":0,"elapsed":2,"user":{"displayName":"Salma Ahmed","userId":"10865891360196912478"}}},"outputs":[],"source":["TOP_TEAMS = [\"Mercedes\" , \"Ferrari\", \"Red Bull\"]\n","\n","\n","OTHER_TEAMS = [\"Sauber\", \"Toro Rosso\", \"Haas F1 Team\", \"Williams\", \"Force India\", \"Renault\", \"AlphaTauri\", \"Aston Martin\", \"McLaren\", \"Alpine\", \"Alpine\", \"Racing Point\", \"Visa Cash App R8 F1 Team\", \"Stake F1 Team Kick Sauber\"]\n","\n","def get_team_tier(team):\n"," if team in TOP_TEAMS:\n"," return \"Top_Tier\"\n"," return \"Other_Tier\""]},{"cell_type":"code","source":["\n","#Section 2-Loading and cleaning the data\n","\n","print(\"Loading data\")\n","\n","LAPS_BEFORE_df = pd.read_csv(DATA_PATH)\n","print(f\"Data loaded: {LAPS_BEFORE_df.shape[0]:,} rows and {LAPS_BEFORE_df.shape[1]} columns.\")\n","\n","print(\"\\nData set columns:\")\n","print(LAPS_BEFORE_df.columns.tolist())\n","\n","# get rid of missing target values\n","before_rows = len(LAPS_BEFORE_df)\n","raw_laps_df = LAPS_BEFORE_df.dropna(subset=[TARGET_COLUMN]).copy()\n","after_rows = len(raw_laps_df)\n","dropped_target = before_rows - after_rows\n","\n","print(f\"\\nRemoved {dropped_target} laps where '{TARGET_COLUMN}' was missing.\")\n","print(f\"Remaining laps: {after_rows:,}\")\n","\n","# pick my features\n","feature_columns = [\"Stint Length\", \"Air_Temp_C\", \"Tire Compound\", \"Constructor\"]\n","\n","print(\"\\nFeatures:\")\n","print(feature_columns)\n","\n","# grab just the columns i need\n","clean_laps_df = raw_laps_df[feature_columns + [TARGET_COLUMN]].copy()\n","\n","# remove rows with missing data\n","clean_laps_df = clean_laps_df.dropna(subset=[\"Stint Length\", \"Air_Temp_C\"])\n","\n","print(\"\\nFirst few rows:\")\n","print(clean_laps_df.head())\n","\n","# add team category\n","team_cats = []\n","for team in clean_laps_df[\"Constructor\"]:\n"," team_cats.append(get_team_tier(team))\n","clean_laps_df[\"Team_Category\"] = team_cats\n","\n","clean_laps_df = clean_laps_df.drop(columns=[\"Constructor\"])\n","\n","# turn categories into numbers\n","laps_encoded = pd.get_dummies(clean_laps_df, columns=[\"Tire Compound\", \"Team_Category\"], drop_first=False)\n","\n","print(\"\\nColumns after encoding:\")\n","print(laps_encoded.columns.tolist())\n","\n","# split into inputs and target\n","feature_columns = []\n","for col in laps_encoded.columns:\n"," if col != TARGET_COLUMN:\n"," feature_columns.append(col)\n","\n","X = laps_encoded[feature_columns]\n","y = laps_encoded[TARGET_COLUMN]\n","\n","print(f\"\\nX has {X.shape[0]:,} rows and {X.shape[1]} features.\")\n","print(f\"y has {len(y):,} values.\")\n","\n","# split into training and test\n","X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.30, random_state=10)\n","\n","print(f\"\\nTraining set: {X_train.shape[0]:,} laps\")\n","print(f\"Test set: {X_test.shape[0]:,} laps\")\n"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"G8Qp5KSu5Km_","executionInfo":{"status":"ok","timestamp":1766863211482,"user_tz":0,"elapsed":551,"user":{"displayName":"Salma Ahmed","userId":"10865891360196912478"}},"outputId":"39372a12-96df-43ba-d614-a2c7210f3cd3"},"execution_count":18,"outputs":[{"output_type":"stream","name":"stdout","text":["Loading data\n","Data loaded: 7,374 rows and 30 columns.\n","\n","Data set columns:\n","['Season', 'Round', 'Circuit', 'Driver', 'Constructor', 'Laps', 'Position', 'TotalPitStops', 'AvgPitStopTime', 'Race Name', 'Date', 'Time_of_race', 'Location', 'Country', 'Air_Temp_C', 'Track_Temp_C', 'Humidity_%', 'Wind_Speed_KMH', 'Lap Time Variation', 'Total Pit Stops', 'Tire Usage Aggression', 'Fast Lap Attempts', 'Position Changes', 'Driver Aggression Score', 'Abbreviation', 'Stint', 'Tire Compound', 'Stint Length', 'Pit_Lap', 'Pit_Time']\n","\n","Removed 185 laps where 'Lap Time Variation' was missing.\n","Remaining laps: 7,189\n","\n","Features:\n","['Stint Length', 'Air_Temp_C', 'Tire Compound', 'Constructor']\n","\n","First few rows:\n"," Stint Length Air_Temp_C Tire Compound Constructor Lap Time Variation\n","0 25.0 15.783333 ULTRASOFT Ferrari 0.001723\n","1 32.0 15.783333 SOFT Ferrari 0.001723\n","2 17.0 15.783333 ULTRASOFT Mercedes 0.001735\n","3 39.0 15.783333 SOFT Mercedes 0.001735\n","4 17.0 15.783333 ULTRASOFT Ferrari 0.001603\n","\n","Columns after encoding:\n","['Stint Length', 'Air_Temp_C', 'Lap Time Variation', 'Tire Compound_HARD', 'Tire Compound_HYPERSOFT', 'Tire Compound_INTERMEDIATE', 'Tire Compound_MEDIUM', 'Tire Compound_SOFT', 'Tire Compound_SUPERSOFT', 'Tire Compound_ULTRASOFT', 'Tire Compound_WET', 'Team_Category_Other_Tier', 'Team_Category_Top_Tier']\n","\n","X has 6,732 rows and 12 features.\n","y has 6,732 values.\n","\n","Training set: 4,712 laps\n","Test set: 2,020 laps\n"]}]},{"cell_type":"code","execution_count":19,"metadata":{"id":"9efd839b-5405-4268-82ad-144177741723","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1766863212465,"user_tz":0,"elapsed":981,"user":{"displayName":"Salma Ahmed","userId":"10865891360196912478"}},"outputId":"dc35d88a-4406-49a4-f222-cf5c5bb7bacf"},"outputs":[{"output_type":"stream","name":"stdout","text":["Completed training\n","Added teams\n","\n","First rows:\n"," Actual Predicted Team_tier\n","6556 0.002354 0.005305 Other_Tier\n","1099 0.002361 0.043686 Top_Tier\n","5403 0.004247 0.005305 Top_Tier\n","820 0.001671 0.005305 Other_Tier\n","6791 0.004027 0.020729 Other_Tier\n","\n","Summary\n","Top tier MAE: 0.031907994174537706\n","Other tier MAE: 0.03495958840016313\n","Average MAE: 0.0339882145946794\n","Bias: 0.003051594225625426\n","Bias percent: 9.127275454339479\n","Moderate bias\n","Worse for other teams\n"]}],"source":["#SECTION 3: Bias proof (unified model)\n","\n","# make model\n","unified_model = DecisionTreeRegressor(random_state=42, max_depth=6, min_samples_leaf=20)\n","unified_model.fit(X_train, y_train)\n","\n","print(\"Completed training\")\n","\n","# predictions\n","preds = unified_model.predict(X_test)\n","\n","# results table\n","results = X_test.copy()\n","results[\"Actual\"] = y_test.values\n","results[\"Predicted\"] = preds\n","\n","# team tier column\n","results[\"Team_tier\"] = np.where(\n"," results[\"Team_Category_Top_Tier\"] == 1,\n"," \"Top_Tier\",\n"," \"Other_Tier\")\n","\n","print(\"Added teams\")\n","\n","print(\"\\nFirst rows:\")\n","print(results[[\"Actual\", \"Predicted\", \"Team_tier\"]].head())\n","\n","# top tier mae\n","top_actual = []\n","top_pred = []\n","for i in results.index:\n"," if results.loc[i, \"Team_tier\"] == \"Top_Tier\":\n"," top_actual.append(results.loc[i, \"Actual\"])\n"," top_pred.append(results.loc[i, \"Predicted\"])\n","\n","top_mae = mean_absolute_error(top_actual, top_pred)\n","\n","# other tier mae\n","other_actual = []\n","other_pred = []\n","for i in results.index:\n"," if results.loc[i, \"Team_tier\"] == \"Other_Tier\":\n"," other_actual.append(results.loc[i, \"Actual\"])\n"," other_pred.append(results.loc[i, \"Predicted\"])\n","\n","other_mae = mean_absolute_error(other_actual, other_pred)\n","\n","# bias\n","diff = top_mae - other_mae\n","if diff < 0:\n"," diff = diff * -1\n","\n","avg = (top_mae + other_mae) / 2\n","pct = (diff / avg) * 100\n","\n","total = mean_absolute_error(results[\"Actual\"], results[\"Predicted\"])\n","\n","print(\"\\nSummary\")\n","print(\"Top tier MAE:\", top_mae)\n","print(\"Other tier MAE:\", other_mae)\n","print(\"Average MAE:\", total)\n","print(\"Bias:\", diff)\n","print(\"Bias percent:\", pct)\n","\n","if pct < 5:\n"," print(\"Very small bias\")\n","if pct >= 5 and pct < 20:\n"," print(\"Moderate bias\")\n","if pct >= 20:\n"," print(\"High bias\")\n","\n","if other_mae > top_mae:\n"," print(\"Worse for other teams\")\n","if top_mae > other_mae:\n"," print(\"Worse for top teams\")\n","if top_mae == other_mae:\n"," print(\"Same for both\")\n","\n","test_results = results.copy()\n","\n","\n","# Unified model MAE calculations (needed for Section 6)\n","top_tier_mae = top_mae # from the unified model results\n","other_tier_mae = other_mae # from the unified model results\n","\n","# Unified model bias\n","bias_amount = abs(top_tier_mae - other_tier_mae)\n","\n"]},{"cell_type":"code","execution_count":20,"metadata":{"id":"018ade54-132e-478d-b866-0d6c52af9cb3","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1766863212561,"user_tz":0,"elapsed":95,"user":{"displayName":"Salma Ahmed","userId":"10865891360196912478"}},"outputId":"d74b0f00-6594-48cf-9bec-62a93ed99fb1"},"outputs":[{"output_type":"stream","name":"stdout","text":["\n","Top-team training rows: 1441\n","Other-team training rows: 3271\n","(643, 13)\n","(1377, 13)\n","----- Specialised Models (Two Models) -----\n","Top Tier MAE (specialised): 0.0362\n","Other Tier MAE (specialised): 0.0360\n"]}],"source":["\n","#Section 4\n","\n","\n","TEAM_COLUMN = \"Team_Category_Top_Tier\" # ← ADD THIS LINE (only new line you need)\n","\n"," #We combine X_train + y_train so we can filter by top teams\n","\n","train_data = pd.concat([X_train, y_train], axis=1) #combines X_train and y_train columns into 1 table side by side (axis=1=side by side column wise, axis=0= row wise)\n","\n","#we are splitting only the training data into top tier and other tier groups using yes for teop teams and no for other teams\n","df_top_train = train_data[train_data[TEAM_COLUMN] == 1 ] #we are filtering the train_data and creating a new daa_table for just top teams and just other teams\n","df_other_train = train_data[train_data[TEAM_COLUMN] == 0]\n","\n","print(\"\\nTop-team training rows:\", df_top_train.shape[0])\n","print(\"Other-team training rows:\", df_other_train.shape[0])\n","\n","#Split each group into new X (inputs) and y (target)\n","\n","#TOP TEAMS\n","X_top_train = df_top_train.drop(columns=[TARGET_COLUMN]) #we drop the lap time variation as this is not an input it is something we want the AI to predict\n","y_top_train = df_top_train[TARGET_COLUMN] #this is the target value we want the AI to predict for our teams\n","\n","\n","\n","X_top_train.head() #shows the data set with the columns selected dropped\n","\n","X_top_train.shape\n","y_top_train.shape\n","\n","\n","#OTHER TEAMS\n","\n","X_other_train= df_other_train.drop(columns=[TARGET_COLUMN]) #we drop the lap time variation as this is not an input it is something we want the AI to predict\n","y_other_train = df_other_train[TARGET_COLUMN] #this is the target value we want the AI to predict for our teams\n","\n","\n","X_other_train.head() #shows the data set with the columns selected dropped\n","X_other_train.shape\n","y_other_train.shape\n","\n","\n","#Train 2 separate models\n","\n","rg1 = DecisionTreeRegressor(max_depth=6, min_samples_leaf=20, random_state=42)\n"," #Model trained ONLY on top teams\n","rg1.fit(X_top_train, y_top_train)\n","\n","rg2 = DecisionTreeRegressor(max_depth=6, min_samples_leaf=20, random_state=42)\n","rg2.fit(X_other_train, y_other_train) #Model trained ONLY on other teams\n","\n","\n","\n","#Prepare test data (combine X_test and y_test)\n","\n","test_data = pd.concat([X_test, y_test], axis=1)\n","\n","\n","\n","\n","#split test data into top and other\n","\n","df_top_test = test_data[test_data[TEAM_COLUMN] == 1]\n","print(df_top_test.shape)\n","df_top_test.head()\n","\n","df_other_test = test_data[test_data[TEAM_COLUMN] == 0]\n","print(df_other_test.shape)\n","df_other_test.head()\n","\n","\n","#prepare x and y for top team test set\n","\n","X_top_test = df_top_test.drop(columns=[TARGET_COLUMN])\n","y_top_test = df_top_test[TARGET_COLUMN]\n","\n","X_top_test.head()\n","X_top_test.shape\n","y_top_test.shape\n","\n","#prepare x and y for other teams test set\n","\n","X_other_test = df_other_test.drop(columns=[TARGET_COLUMN])\n","y_other_test = df_other_test[TARGET_COLUMN]\n","\n","X_other_test.head()\n","X_other_test.shape\n","y_other_test.shape\n","\n","\n","#Predict using the trained models\n","\n","y_predict_top = rg1.predict(X_top_test)\n","\n","y_predict_other = rg2.predict(X_other_test)\n","\n","\n","# SPECIALISED model errors (two-model setup)\n","top_tier_mae_specialised = mean_absolute_error(y_top_test, y_predict_top)\n","other_tier_mae_specialised = mean_absolute_error(y_other_test, y_predict_other)\n","\n","print(\"----- Specialised Models (Two Models) -----\")\n","print(f\"Top Tier MAE (specialised): {top_tier_mae_specialised:.4f}\")\n","print(f\"Other Tier MAE (specialised): {other_tier_mae_specialised:.4f}\")"]},{"cell_type":"code","execution_count":21,"metadata":{"id":"b0173ae6-1dde-47a8-9c0e-29090b8ef31d","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1766863212566,"user_tz":0,"elapsed":3,"user":{"displayName":"Salma Ahmed","userId":"10865891360196912478"}},"outputId":"bc46e18b-ac1b-412e-8b4a-2a86327b8197"},"outputs":[{"output_type":"stream","name":"stdout","text":["New Bias: 0.00023884712935314661\n","New Bias (two models): 0.0002\n","Specialised model predicts better for OTHER teams.\n"]}],"source":["# SECTION 5 — Mitigation evaluation (specialised model bias)\n","\n","new_bias = abs(top_tier_mae_specialised - other_tier_mae_specialised)\n","\n","print(\"New Bias:\", new_bias)\n","print(f\"New Bias (two models): {new_bias:.4f}\")\n","\n","if top_tier_mae_specialised < other_tier_mae_specialised:\n"," print(\"Specialised model predicts better for TOP teams.\")\n","elif top_tier_mae_specialised > other_tier_mae_specialised:\n"," print(\"Specialised model predicts better for OTHER teams.\")\n","else:\n"," print(\"Both groups have equal error — perfectly fair model.\")\n","\n","\n"]},{"cell_type":"code","source":["#Section 6\n","\n","print(\"\\n\")\n","\n","print(\"Section 6: compare one model vs two models\")\n","\n","print(\"\\n\")\n","\n","\n","# 1. show a simple comparison table\n","\n","groups = [\"Top_Tier\", \"Other_Tier\"]\n","\n","unified_mae_list = [top_tier_mae, other_tier_mae]\n","\n","special_mae_list = [top_tier_mae_specialised, other_tier_mae_specialised]\n","\n","\n","print(\"Group Unified_MAE Specialised_MAE\")\n","\n","print(groups[0], \" \", round(unified_mae_list[0], 4), \" \", round(special_mae_list[0], 4))\n","\n","print(groups[1], \" \", round(unified_mae_list[1], 4), \" \", round(special_mae_list[1], 4))\n","\n","\n","# 2. work out bias before and after\n","\n","old_bias = bias_amount\n","\n","new_bias = top_tier_mae_specialised - other_tier_mae_specialised\n","\n","if new_bias < 0:\n","\n"," new_bias = new_bias * -1\n","\n","\n","# 3. percentage change in bias\n","\n","bias_change_pct = 0\n","\n","if old_bias != 0:\n","\n"," bias_change_pct = (old_bias - new_bias) / old_bias\n","\n"," bias_change_pct = bias_change_pct * 100\n","\n","\n","print(\"\\nOld bias (unified model):\", old_bias)\n","\n","print(\"New bias (specialised models):\", new_bias)\n","\n","print(\"Bias change:\", round(bias_change_pct, 2), \"%\")\n","\n","\n","# 4. say if it got better or worse\n","\n","if new_bias < old_bias:\n","\n"," print(\"Specialised models reduced the bias\")\n","\n","elif new_bias > old_bias:\n","\n"," print(\"Specialised models increased the bias\")\n","\n","else:\n","\n"," print(\"Bias stayed the same\")\n","\n","\n","# positions for the two groups on x-axis\n","\n","pos_top = 0\n","\n","pos_other = 1\n","\n","width = 0.35\n","\n","\n","# positions for unified and specialised bars\n","\n","pos_top_unified = pos_top - width/2\n","\n","pos_top_special = pos_top + width/2\n","\n","pos_other_unified = pos_other - width/2\n","\n","pos_other_special = pos_other + width/2\n","\n","\n","plt.figure()\n","\n","\n","# draw each bar\n","\n","plt.bar(pos_top_unified, unified_mae_list[0], width, label=\"Unified\",color=\"tab:blue\")\n","\n","plt.bar(pos_top_special, special_mae_list[0], width, label=\"Specialised\", color=\"tab:orange\")\n","\n","plt.bar(pos_other_unified, unified_mae_list[1], width,color=\"tab:blue\")\n","\n","plt.bar(pos_other_special, special_mae_list[1], width,color=\"tab:orange\")\n","\n","\n","plt.xticks([pos_top, pos_other], groups)\n","\n","plt.xlabel(\"Group\")\n","\n","plt.ylabel(\"MAE\")\n","\n","plt.title(\"Unified vs Specialised models\")\n","\n","\n","# add MAE numbers on top of bars\n","\n","plt.text(pos_top_unified, unified_mae_list[0], str(round(unified_mae_list[0], 3)), ha='center')\n","\n","plt.text(pos_top_special, special_mae_list[0], str(round(special_mae_list[0], 3)), ha='center')\n","\n","plt.text(pos_other_unified, unified_mae_list[1], str(round(unified_mae_list[1], 3)), ha='center')\n","\n","plt.text(pos_other_special, special_mae_list[1], str(round(special_mae_list[1], 3)), ha='center')\n","\n","\n","plt.legend()\n","\n","plt.show()"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":704},"id":"1mt1U1SxBuCJ","executionInfo":{"status":"ok","timestamp":1766863213503,"user_tz":0,"elapsed":937,"user":{"displayName":"Salma Ahmed","userId":"10865891360196912478"}},"outputId":"cc17f8ee-033c-48df-acb6-f2e1f43cb7b7"},"execution_count":22,"outputs":[{"output_type":"stream","name":"stdout","text":["\n","\n","Section 6: compare one model vs two models\n","\n","\n","Group Unified_MAE Specialised_MAE\n","Top_Tier 0.0319 0.0362\n","Other_Tier 0.035 0.036\n","\n","Old bias (unified model): 0.003051594225625426\n","New bias (specialised models): 0.00023884712935314661\n","Bias change: 92.17 %\n","Specialised models reduced the bias\n"]},{"output_type":"display_data","data":{"text/plain":["
"],"image/png":"iVBORw0KGgoAAAANSUhEUgAAAkgAAAHHCAYAAABEEKc/AAAAOnRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjEwLjAsIGh0dHBzOi8vbWF0cGxvdGxpYi5vcmcvlHJYcgAAAAlwSFlzAAAPYQAAD2EBqD+naQAAWlNJREFUeJzt3XtcjvfjP/DXXXTfKpXOB6kQOUSUDuyzmIgZGpPTZxFjM+fMyFBmlo0sm7a+Ydg+iAyfOXxMC3MokXKaw5yjo1hF0fH6/eHXNffVHUV1F6/n43E/uN/X+3pf7+u+uXr1vt7XdckEQRBARERERCINdXeAiIiIqL5hQCIiIiKSYEAiIiIikmBAIiIiIpJgQCIiIiKSYEAiIiIikmBAIiIiIpJgQCIiIiKSYEAiIiIikmBAIqonxo4dC1tbW6Wyhw8f4oMPPoC5uTlkMhlmzJiBmzdvQiaTYf369TW27fXr10Mmk+HmzZs11uar6NChQ5DJZDh06FC11w0ODoZMJlMqs7W1xdixY2umc1XUUL7rl/msG8o+Uv3GgERUDeU/5LKzs1Uu79ixI3r27Flj2/vyyy+xfv16TJo0CT///DPef//9Gmu7vtq1axc8PT1hamoKbW1ttGzZEr6+vti3b5+6u0ZEr5FG6u4AET2xevVqlJWVKZUdOHAA7u7uCAoKEssEQcCjR4/QuHHjuu5irVu+fDlmz54NT09PBAYGQltbG1evXsXvv/+OqKgo9OvXT639e/PNN/Ho0SNoaWnVSHuXL1+GhgZ/TyWqjxiQiOoJVYEnKysL7du3VyqTyWRQKBR11a06U1JSgsWLF6NPnz7Yv39/heVZWVlq6JUyDQ2NGv3s5XJ5jbVFRDWLv7oQ1aLyeRRbt27FkiVL0Lx5cygUCvTu3RtXr15Vqvv0HKTy9W7cuIE9e/ZAJpOJcyoqm4N06dIlvPfeezA0NIRCoYCLiwt+/fXXCn36888/8dZbb6FJkyZo3rw5vvjiiwojV6osX74cMpkMt27dqrAsMDAQWlpa+PvvvwEAV65cwdChQ2Fubg6FQoHmzZtjxIgRyM3NrbT97Oxs5OXloUePHiqXm5qain8v/3y2bNmCefPmwdzcHDo6Ohg0aBBu375dYd2EhAT069cP+vr60NbWhqenJ44dO1ahXmpqKsaPHw9LS0vI5XLY2dlh0qRJKCoqUtru0/Nijhw5gmHDhqFFixaQy+WwtrbGzJkz8ejRo0r3tZx0DlJxcTEWLVoEe3t7KBQKGBkZ4Y033kBMTIzSerX9XQNP/j3q6uoiJSUF77zzDnR1dWFlZYXw8HAAwLlz5/DWW29BR0cHNjY22LRpU4U2rl+/jmHDhsHQ0BDa2tpwd3fHnj17KtS7c+cOfHx8oKOjA1NTU8ycOROFhYUq+1XV71IqMTER3t7eMDY2RpMmTWBnZ4dx48ZV6bOg1xNHkIjqwNKlS6GhoYFPPvkEubm5+PrrrzF69GgkJCSorN+uXTv8/PPPmDlzJpo3b45Zs2YBAExMTHD37t0K9f/880/06NEDVlZWmDt3LnR0dLB161b4+Pjgl19+wbvvvgsAyMjIQK9evVBSUiLWi4yMRJMmTZ67D76+vvj000+xdetWzJ49W2nZ1q1b0bdvXzRr1gxFRUXw9vZGYWEhpk6dCnNzc6SmpmL37t3IycmBvr6+yvZNTU3RpEkT7Nq1C1OnToWhoeFz+7RkyRLIZDLMmTMHWVlZCAsLg5eXF06fPi3u04EDB9C/f384OzsjKCgIGhoaWLduHd566y0cOXIErq6uAIC0tDS4uroiJycHEydOhIODA1JTU7Ft2zYUFBRUelotOjoaBQUFmDRpEoyMjHDixAl89913uHPnDqKjo5+7D08LDg5GSEgIPvjgA7i6uiIvLw+JiYlISkpCnz59ANTNd12utLQU/fv3x5tvvomvv/4aGzduxJQpU6Cjo4PPPvsMo0ePxpAhQxAREQE/Pz94eHjAzs4OAJCZmYnu3bujoKAA06ZNg5GRETZs2IBBgwZh27ZtYj8fPXqE3r17IyUlBdOmTYOlpSV+/vlnHDhwoEJ/qvpdSmVlZaFv374wMTHB3LlzYWBggJs3b2L79u3V+n7oNSMQUZUFBQUJAIS7d++qXN6hQwfB09NTfH/w4EEBgNCuXTuhsLBQLF+5cqUAQDh37pxYNmbMGMHGxkapPRsbG2HAgAFKZTdu3BAACOvWrRPLevfuLTg6OgqPHz8Wy8rKyoTu3bsL9vb2YtmMGTMEAEJCQoJYlpWVJejr6wsAhBs3bjxz/z08PARnZ2elshMnTggAhJ9++kkQBEFITk4WAAjR0dHPbEuVhQsXCgAEHR0doX///sKSJUuEU6dOVahX/rlaWVkJeXl5YvnWrVsFAMLKlSvFz8De3l7w9vYWysrKxHoFBQWCnZ2d0KdPH7HMz89P0NDQEE6ePFlhe+Xrlm/34MGDSm1JhYSECDKZTLh165ZYVv5v52k2NjbCmDFjxPedO3eu8H1L1dV3PWbMGAGA8OWXX4plf//9t9CkSRNBJpMJUVFRYvmlS5cEAEJQUFCF7R85ckQse/DggWBnZyfY2toKpaWlgiAIQlhYmABA2Lp1q1gvPz9faN26tdJnXZ3vct26dUr7uGPHDgGAyu+WqDI8xUZUB/z9/ZVGIP71r38BeHIK4mXdv38fBw4cgK+vLx48eIDs7GxkZ2fj3r178Pb2xpUrV5CamgoA2Lt3L9zd3ZV+0zYxMcHo0aOrtK3hw4fj1KlTuHbtmli2ZcsWyOVyDB48GADEEaLffvsNBQUF1dqXRYsWYdOmTejSpQt+++03fPbZZ3B2dkbXrl1x8eLFCvX9/PzQtGlT8f17770HCwsL7N27FwBw+vRpXLlyBaNGjcK9e/fEzyY/Px+9e/fG4cOHUVZWhrKyMuzcuRMDBw6Ei4tLhe1IL89/2tMjMvn5+cjOzkb37t0hCAKSk5Ortf8GBgb4888/ceXKFZXL6/K7LvfBBx8o9a9t27bQ0dGBr6+vWN62bVsYGBgo/Xveu3cvXF1d8cYbb4hlurq6mDhxIm7evIkLFy6I9SwsLPDee++J9bS1tTFx4kSlflT1u1TFwMAAALB7924UFxdXa//p9cWARFTDVP0wbdGihdL7Zs2aAYA4Z+dlXL16FYIgYMGCBTAxMVF6lV/9Vj7B+datW7C3t6/QRtu2bau0rWHDhkFDQwNbtmwB8OSKuujoaPTv3x96enoAADs7OwQEBGDNmjUwNjaGt7c3wsPDnzn/6GkjR47EkSNH8Pfff2P//v0YNWoUkpOTMXDgQDx+/FiprnRfZDIZWrduLd7/pjxojBkzpsJns2bNGhQWFiI3Nxd3795FXl4eOnbsWKU+Pi0lJQVjx46FoaEhdHV1YWJiAk9PTwCo8j6X+/zzz5GTk4M2bdrA0dERs2fPxtmzZ8XldfldA4BCoYCJiYlSmb6+Ppo3b17h37m+vr7Sv+dbt26p3Fa7du3E5eV/tm7dukJ70nWr+l2q4unpiaFDh2LRokUwNjbG4MGDsW7dukrnOREBnINEVC3lVzBVNgG3oKBA5VVOmpqaKusLgvDSfSr/rfmTTz6Bt7e3yjqtW7d+6e0AgKWlJf71r39h69atmDdvHo4fP46UlBR89dVXSvVCQ0MxduxY/Pe//8X+/fsxbdo0hISE4Pjx42jevHmVtqWnp4c+ffqgT58+aNy4MTZs2ICEhAQxfFRF+WezbNkyODk5qayjq6uL+/fvV7nNp5WWlqJPnz64f/8+5syZAwcHB+jo6CA1NRVjx46t8oTocm+++SauXbsmfm5r1qzBN998g4iICHzwwQd1+l0Dlf+7rc1/z5Wp6nepikwmw7Zt23D8+HHs2rULv/32G8aNG4fQ0FAcP3680vXo9caARFQNNjY2AJ7cv8ba2lppWUFBAW7fvo2+ffvWaZ9atmwJ4MltAry8vJ5Z18bGRuXpm8uXL1d5e8OHD8fHH3+My5cvY8uWLdDW1sbAgQMr1HN0dISjoyPmz5+PuLg49OjRAxEREfjiiy+qvK1yLi4u2LBhA9LT05XKpfsiCAKuXr2KTp06AQBatWoF4EnYetZnY2JiAj09PZw/f75a/Tp37hz++usvbNiwAX5+fmK59Kqz6jA0NIS/vz/8/f3x8OFDvPnmmwgODsYHH3xQ59/1y7CxsVG5rUuXLonLy/88f/48BEFQGkWSrlvV7/JZ3N3d4e7ujiVLlmDTpk0YPXo0oqKilE4jEpXjKTaiaujduze0tLTwww8/VBgdiIyMRElJCfr371+nfTI1NUXPnj3xf//3fxUCBAClq97efvttHD9+HCdOnFBavnHjxipvb+jQodDU1MTmzZsRHR2Nd955Bzo6OuLyvLw8lJSUKK3j6OgIDQ2NZ57SKCgoQHx8vMpl//vf/wBUPO3y008/4cGDB+L7bdu2IT09XfwOnJ2d0apVKyxfvhwPHz6s0G75Z6OhoQEfHx/s2rULiYmJFepVNjJSPpLy9HJBELBy5cpK9/NZ7t27p/ReV1cXrVu3Fj+3uv6uX8bbb7+NEydOKH2n+fn5iIyMhK2trXh/r7fffhtpaWnYtm2bWK+goACRkZFK7VX1u1Tl77//rvAdlo9C8TQbVYYjSETVYGpqioULF2L+/Pl48803MWjQIGhrayMuLg6bN29G3759VY6m1Lbw8HC88cYbcHR0xIQJE9CyZUtkZmYiPj4ed+7cwZkzZwAAn376KX7++Wf069cP06dPFy/9trGxUZrr8iympqbo1asXVqxYgQcPHmD48OFKyw8cOIApU6Zg2LBhaNOmDUpKSvDzzz9DU1MTQ4cOrbTdgoICdO/eHe7u7ujXrx+sra2Rk5ODnTt34siRI/Dx8UGXLl2U1jE0NMQbb7wBf39/ZGZmIiwsDK1bt8aECRMAPAk+a9asQf/+/dGhQwf4+/vDysoKqampOHjwIPT09LBr1y4ATx7rsn//fnh6emLixIlo164d0tPTER0djaNHj4oTfZ/m4OCAVq1a4ZNPPkFqair09PTwyy+/vPDcsvbt26Nnz55wdnaGoaEhEhMTsW3bNkyZMkWsU5ff9cuYO3cuNm/ejP79+2PatGkwNDTEhg0bcOPGDfzyyy/iHcQnTJiAVatWwc/PD6dOnYKFhQV+/vlnaGtrK7VXne9SasOGDfj+++/x7rvvolWrVnjw4AFWr14NPT09vP3227X+WVADpZ6L54gatv/85z+Cu7u7oKOjI8jlcsHBwUFYtGiR0qXXgvDPZeHSS95VXar/Mpf5C4IgXLt2TfDz8xPMzc2Fxo0bC1ZWVsI777wjbNu2Tane2bNnBU9PT0GhUAhWVlbC4sWLhbVr11bp0u9yq1evFgAITZs2FR49eqS07Pr168K4ceOEVq1aCQqFQjA0NBR69eol/P77789ss7i4WFi9erXg4+Mj2NjYCHK5XNDW1ha6dOkiLFu2TOk2CeWf6+bNm4XAwEDB1NRUaNKkiTBgwAClS+vLJScnC0OGDBGMjIwEuVwu2NjYCL6+vkJsbKxSvVu3bgl+fn6CiYmJIJfLhZYtWwqTJ08Wt63qMv8LFy4IXl5egq6urmBsbCxMmDBBOHPmTIXvqCqX+X/xxReCq6urYGBgIDRp0kRwcHAQlixZIhQVFSmtVxff9ZgxYwQdHZ0K5Z6enkKHDh0qlKv6t3rt2jXhvffeEwwMDASFQiG4uroKu3fvrrDurVu3hEGDBgna2tqCsbGxMH36dGHfvn0VPmtBqNp3Kb3MPykpSRg5cqTQokULQS6XC6ampsI777wjJCYmPvMzoNebTBBqcVYdEVEtOHToEHr16oXo6Gily8OJiGoK5yARERERSTAgEREREUkwIBERERFJcA4SERERkQRHkIiIiIgkGJCIiIiIJHijyBdUVlaGtLQ0NG3a9JlP+iYiIqL6QxAEPHjwAJaWluINS1VhQHpBaWlpFZ7FRURERA3D7du3n/nwbAakF9S0aVMATz5gPT09NfeGiIiIqiIvLw/W1tbiz/HKMCC9oPLTanp6egxIREREDczzpsdwkjYRERGRBAMSERERkQQDEhEREZEE5yDVstLSUhQXF6u7G1RHGjduDE1NTXV3g4iIXhIDUi0RBAEZGRnIyclRd1eojhkYGMDc3Jz3xyIiasAYkGpJeTgyNTWFtrY2f1i+BgRBQEFBAbKysgAAFhYWz10nPDwcy5YtQ0ZGBjp37ozvvvsOrq6uldaPjo7GggULcPPmTdjb2+Orr77C22+/LS4PDg5GVFQUbt++DS0tLTg7O2PJkiVwc3NTamfPnj34/PPPcfbsWSgUCnh6emLnzp0vtuNERK8gBqRaUFpaKoYjIyMjdXeH6lCTJk0AAFlZWTA1NX3m6bYtW7YgICAAERERcHNzQ1hYGLy9vXH58mWYmppWqB8XF4eRI0ciJCQE77zzDjZt2gQfHx8kJSWhY8eOAIA2bdpg1apVaNmyJR49eoRvvvkGffv2xdWrV2FiYgIA+OWXXzBhwgR8+eWXeOutt1BSUoLz58/XwqdBRNSACfRCcnNzBQBCbm5uhWWPHj0SLly4IBQUFKihZ6RuBQUFwoULF4RHjx49s56rq6swefJk8X1paalgaWkphISEqKzv6+srDBgwQKnMzc1N+PDDDyvdRvm/099//10QBEEoLi4WrKyshDVr1lR1d4hIjVatWiXY2NgIcrlccHV1FRISEp5Zf+vWrULbtm0FuVwudOzYUdizZ4/S8qCgIKFt27aCtra2YGBgIPTu3Vs4fvx4hXZ2794tuLq6CgqFQjAwMBAGDx5ck7ulVs/6+f00XsVWi3ha7fVUle+9qKgIp06dgpeXl1imoaEBLy8vxMfHq1wnPj5eqT4AeHt7V1q/qKgIkZGR0NfXR+fOnQEASUlJSE1NhYaGBrp06QILCwv079+fI0hE9VD5KHNQUBCSkpLQuXNneHt7i6fxpcpHmcePH4/k5GT4+PjAx8dH6f93+SjzuXPncPToUdja2qJv3764e/euWOeXX37B+++/D39/f5w5cwbHjh3DqFGjan1/6506CmyvnKqMID1vBIFeTVX5/lNTUwUAQlxcnFL57NmzBVdXV5XrNG7cWNi0aZNSWXh4uGBqaqpUtmvXLkFHR0eQyWSCpaWlcOLECXHZ5s2bBQBCixYthG3btgmJiYnCyJEjBSMjI+HevXvV3VUiqkUcZa4dHEEitbG1tUVYWJj4PiMjA3369IGOjg4MDAwAPBlledlJwWPHjoWPj89LtfEq6tWrF06fPo24uDj069cPvr6+4m+cZWVlAIDPPvsMQ4cOhbOzM9atWweZTIbo6Gh1dpuInsJRZvXjJO06ZDt3T51u7+bSAdWq37NnTzg5OSmFGwBYv349ZsyYUeVbFpw8eRI6Ojri+2+++Qbp6ek4ffo09PX1AQDp6elo1qxZtfr3KjE2NoampiYyMzOVyjMzM2Fubq5yHXNz8yrV19HRQevWrdG6dWu4u7vD3t4ea9euRWBgoHhlXfv27cX6crkcLVu2REpKSk3sGhHVgOzsbJSWlsLMzEyp3MzMDJcuXVK5TkZGhsr6GRkZSmW7d+/GiBEjUFBQAAsLC8TExMDY2BgAcP36dQBProhdsWIFbG1tERoaip49e+Kvv/6CoaFhTe1ivccRJKpxJiYm0NbWFt9fu3YNzs7OsLe3F6/OMjc3h1wuV1cX1a78EvzY2FixrKysDLGxsfDw8FC5joeHh1J9AIiJiam0/tPtFhYWAgCcnZ0hl8tx+fJlcXlxcTFu3rwJGxubF90dImpAOMpcNQxIVC3lp7WWL18OCwsLGBkZYfLkyUp3C3/6FJutrS1++eUX/PTTT5DJZBg7diyAiqfYbt++DV9fXxgYGMDQ0BCDBw/GzZs3xeWlpaUICAiAgYEBjIyM8Omnn0IQhDrY49oTEBCA1atXY8OGDbh48SImTZqE/Px8+Pv7AwD8/PwQGBgo1p8+fTr27duH0NBQXLp0CcHBwUhMTMSUKVMAAPn5+Zg3bx6OHz+OW7du4dSpUxg3bhxSU1MxbNgwAICenh4++ugjBAUFYf/+/bh8+TImTZoEAGIdIlK/uhhldnd3x9q1a9GoUSOsXbsWADjK/BQGJKq2gwcP4tq1azh48CA2bNiA9evXY/369Srrnjx5UvwNJT09HStXrqxQp7i4GN7e3mjatCmOHDmCY8eOQVdXF/369UNRUREAIDQ0FOvXr8ePP/6Io0eP4v79+9ixY0dt7matGz58OJYvX46FCxfCyckJp0+fxr59+8Qh8pSUFKSnp4v1u3fvjk2bNiEyMhKdO3fGtm3bsHPnTvEeSJqamrh06RKGDh2KNm3aYODAgbh37x6OHDmCDh06iO0sW7YMI0aMwPvvv49u3brh1q1bOHDgwGt9ypOovuEos/pxDhJVW7NmzbBq1SpoamrCwcEBAwYMQGxsLCZMmFChromJCeRyOZo0aVLpbz1btmxBWVkZ1qxZI14iv27dOhgYGODQoUPo27cvwsLCEBgYiCFDhgAAIiIi8Ntvv9XcTqUl11xbJQKQcxdYNQx4ePuZVacAmOIPAAoAl4D/9QX+92TZoZ4AkAwE/xMEhwEYNrK8/m3gxEjgxJNlCgDbOwHoVP4uH8AhYM8h4Knpb40BLNcFlk8CABmAE0B0d6CuR8+Dc+t4g0QNS0BAAMaMGQMXFxe4uroiLCyswiizlZUVQkJCADwZZfb09ERoaCgGDBiAqKgoJCYmIjIyEsCTUeYlS5Zg0KBBsLCwQHZ2NsLDwysdZba2toaNjQ2WLVsG4PUbZWZAomrr0KGD0h2iLSwscO7cuRdu78yZM7h69SqaNm2qVP748WNcu3YNubm5SE9PV3pcRqNGjeDi4tLgT7MREVVm+PDhuHv3LhYuXIiMjAw4OTlVGGXW0PjnRFD5KPP8+fMxb9482Nvbqxxl3rBhA7Kzs2FkZIRu3bqpHGVu1KgR3n//fTx69Ahubm6v5SgzAxKJ9PT0kJtb8bf6nJwc8eoz4MkT658mk8nEiX0v4uHDh3B2dsbGjRsrLCt/PAYRqZ86nh1oa2uLW7duKbUbEhKCuXPn1vwOPkuw/vPr1AKOMqsP5yCRqG3btkhKSqpQnpSUhDZt2tTadrt27YorV67A1NRUvDy9/KWvrw99fX1YWFggISFBXKekpASnTp2qtT4RkTJ13dUZAD7//HOkp6eLr6lTp9bqvhIBDEj0lEmTJuGvv/7CtGnTcPbsWVy+fBkrVqzA5s2bMWvWrFrb7ujRo2FsbIzBgwfjyJEjuHHjBg4dOoRp06bhzp07AJ6cW1+6dCl27tyJS5cu4eOPP67yfZmI6OWtWLECEyZMgL+/P9q3b4+IiAhoa2vjxx9/VFl/5cqV6NevH2bPno127dph8eLF6Nq1K1atWiXWGTVqFLy8vNCyZUt06NABK1asQF5eHs6ePavUVtOmTWFubi6+nr7PGlFt4Sm2OlTdGzfWtZYtW+Lw4cP47LPP4OXlhaKiIjg4OCA6Ohr9+vWrte1qa2vj8OHDmDNnDoYMGYIHDx7AysoKvXv3hp6eHgBg1qxZSE9Px5gxY6ChoYFx48bh3XffVXlKkIhqVvldnZ++7URV7uocEBCgVObt7V3pHfRV3dW53NKlS7F48WK0aNECo0aNwsyZM9GoEX98Ue3ivzBS0q1bN+zfv7/S5aou55feefvp+xcBUHlAlE6uNjc3x4YNGyrdbqNGjRAWFlZhW0RU+9R1V2cAmDZtGrp27QpDQ0PExcUhMDAQ6enpWLFiRQ3tHZFq9eIUW3h4OGxtbaFQKODm5oYTJ048s350dDQcHBygUCjg6OiIvXv3Ki0PDg6Gg4MDdHR00KxZM3h5eSnNXwGeTPyTyWRKr6VLl9b4vhERUeWedVdn4Mml7j179kSnTp3w0UcfITQ0FN9995143x6i2qL2gMSJf0RE9Zu67uqsipubG0pKSiqMVBPVNLUHJE78IyKq39R1V2dVTp8+DQ0NDfG5jkS1Ra0BqXzin5eXl1hWlYl/T9cHnkz8q6z+8yb+GRkZoUuXLli2bBlKSkoq7WthYSHy8vKUXkRErwt1PDswPj4eYWFhOHPmDK5fv46NGzdi5syZ+Pe///3a3bSQ6p5aJ2k3pIl/ISEhWLRo0YvsJhFRg6eOuzrL5XJERUUhODgYhYWFsLOzw8yZMytcHUdUG2SCGp/VkJaWBisrK8TFxSkNu3766af4448/KkysBp4M9W7YsAEjR44Uy77//nssWrRI6Xx3fn4+0tPTkZ2djdWrV+PAgQNISEiodFj2xx9/xIcffoiHDx9CLpdXWF5YWKg07JuXlwdra2vk5uaKl6KXe/z4MW7cuAE7OzsoFIqqfyCkPjX4LLbHJQJupN6F3bFZUDznWWyvNT6L7aXYzt3z/EpUY24qRqm7C6+fWjpG5OXlQV9fX+XP76ep9RRbQ5r4J5fLoaenp/QiIiKiV5NaAxIn/hEREVF9pPYbRQYEBGDMmDFwcXGBq6srwsLCKkz8s7KyQkhICIAnE/88PT0RGhqKAQMGICoqComJiYiMjATw5NTakiVLMGjQIFhYWCA7Oxvh4eEVJv4lJCSgV69eaNq0KeLj4znxj4iIiERqv8x/+PDhWL58ORYuXAgnJyecPn26wsS/9PR0sX75xL/IyEh07twZ27ZtUznxb+jQoWjTpg0GDhyIe/fuqZz45+npiQ4dOmDJkiWYOXOmGLJIvWQyWaWPI1AlODgYTk5O4vuxY8fCx8enxvtV7ubtNMisuuL0+cu1tg0iIlIvtY8gAcCUKVPESz+lDh06VKFs2LBh4miQlEKhwPbt25+5va5du+L48ePV7udLC9av4+1Vf4Jb+VUqe/bsQWZmJpo1a4bOnTtj4cKF6NGjRy10sqL09PSXGslbuXJlhUeZEBERVUe9CEhUfwwdOhRFRUXYsGEDWrZsiczMTMTGxuLevXt11ofKJuhXlb5+HQdRIiJ65aj9FBvVHzk5OThy5Ai++uor9OrVCzY2NnB1dUVgYCAGDRoE4Mnprx9++AH9+/dHkyZN0LJlS2zbtk2pndu3b8PX1xcGBgYwNDTE4MGDK1wd+OOPP6JDhw6Qy+WwsLBQGkGUnmKbM2cO2rRpA21tbbRs2RILFixAcXFxpfshPcW2bds2ODo6okmTJjAyMoKXlxfy8/PF5WvWrEE7zyFQtHSHw5tD8P36rUrtnUg+jy59R0LR0h0u/Ucj+bzqe3QREdGrgwGJRLq6utDV1cXOnTufecXfggULMHToUJw5cwajR4/GiBEjcPHiRQBAcXExvL290bRpUxw5cgTHjh2Drq4u+vXrh6KiIgDADz/8gMmTJ2PixIk4d+4cfv31V7Ru3brS7TVt2hTr16/HhQsXsHLlSqxevRrffPNNlfYpPT0dI0eOxLhx43Dx4kUcOnQIQ4YMEU/Bbdy4EQsXLsSSOZNx8dAv+HLuZCxY9gM2bN0FAHiYX4B3xkxH+zYtcep/GxEc8CE+WRxWpW0TEVHDxVNsJGrUqBHWr1+PCRMmICIiAl27doWnpydGjBiBTp06ifWGDRuGDz74AACwePFixMTE4LvvvsP333+PLVu2oKysDGvWrIFMJgMArFu3DgYGBjh06BD69u2LL774ArNmzcL06dPFNrt161Zpv+bPny/+3dbWFp988gmioqLw6aefPnef0tPTUVJSgiFDhsDGxgYA4OjoKC4PCgpCaGgohng6AADsWljhwl838H//+QVjfAdi047/oaxMwNrlC6FQyNGhbSvcSc/CpMAvq/KREhFRA8WAREqGDh2KAQMG4MiRIzh+/Dj+97//4euvv8aaNWswduxYAKhwzykPDw+cPn0aAHDmzBlcvXoVTZs2Varz+PFjXLt2DVlZWUhLS0Pv3r2r3KctW7bg22+/xbVr1/Dw4UOUlJRU+UadnTt3Ru/eveHo6Ahvb2/07dsX7733Hpo1a4b8/Hxcu3YN48ePxwTZP+uUlJZCv6kuAODilRvo1K41FIp/7q7u4dxJuhkiInrFMCBRBQqFAn369EGfPn2wYMECfPDBBwgKChID0rM8fPgQzs7O2LhxY4VlJiYmSs9qqor4+HiMHj0aixYtgre3N/T19REVFYXQ0NAqra+pqYmYmBjExcVh//79+O677/DZZ58hISEB2traAIDVq1fDzVa3wnpERPT64hwkeq727dsrTWqW3iLh+PHjaNeuHYAnt1C4cuUKTE1N0bp1a6WXvr4+mjZtCltb2wp3Q69MXFwcbGxs8Nlnn8HFxQX29va4detWtfovk8nQo0cPLFq0CMnJydDS0sKOHTtgZmYGS0tLXL9+Ha3tWii97FpYAQDa2dvh7MWrePz4nzlZx5POVmv7RETU8DAgkejevXt466238J///Adnz57FjRs3EB0dja+//hqDBw8W60VHR+PHH3/EX3/9haCgIJw4cUK8Cm306NEwNjbG4MGDceTIEdy4cQOHDh3CtGnTcOfOHQBPbuwYGhqKb7/9FleuXEFSUhK+++47lX2yt7dHSkoKoqKicO3aNXz77bfYsWNHlfcpISEBX375JRITE5GSkoLt27fj7t27YqBbtGgRQkJC8O3azfjr2i2cu3gF67b8Fyv+7z8AgFHv9odMBkyYvRgX/rqOvbFHsTzi5xf6fImIqOHgKba6VM+fXq6rqws3Nzd88803uHbtGoqLi2FtbY0JEyZg3rx5Yr1FixYhKioKH3/8MSwsLLB582a0b98eAKCtrY3Dhw9jzpw5GDJkCB48eAArKyv07t1bnDc0ZswYPH78GN988w0++eQTGBsb47333lPZp0GDBmHmzJmYMmUKCgsLMWDAACxYsADBwcFV2ic9PT0cPnwYYWFhyMvLg42NDUJDQ9G/f38AwAcffABtbW0s+/JzzP4iDDraTeDo0BozPnjy5G5dHW3sWh+Gj+Z+iS7eI9HeviW++mwahk6Y/aIfMxERNQAygbccfiF5eXnQ19dHbm5uhQnDjx8/xo0bN2BnZweFQqGmHtYOmUyGHTt21OqjPNQiLbnGmnpcIuBG6l3YHZsFxcPbNdbuK6ee/8JQ39nO3aPuLrxWbipGqbsLr59aOkY86+f303iKjYiIiEiCAYmIiIhIgnOQqFp4RpaIiF4HHEEiIiIikuAIUi1qaKMtWVlZyMjIQHFxMbS1tdGiRQvo6OhUWv/+/ftIS0tDYWEhFAoFmjdvDn19fXF5Wloa7t+/j6KiIshkMmhra8PKygq6uk9uylhYWIj09HTk5eWhuLgYWlpaMDQ0hIWFRbVvKFmfiF97A/v+iYjoHw33p1A91rhxYwBAQUGBmntSdffv38ft27dhaWmJ9u3bo0mTJvjrr79QXFyssv7Dhw9x/fp1GBsbo3379jAwMMDVq1fx6NEjsY5cLkeLFi3QoUMHODg4QC6X48qVK2Kbjx8/hiAIsLGxQceOHWFtbY27d+8iNTW1Tva5thQUAygtQuPH99TdFSIiekEcQaoFmpqaMDAwQFZWFoAn9wYqf3BrfZWeno5mzZqJoztmZmbIyclBeno6TE1NK9RPS0uDrq4uDAwMAABGRkbIyclBWloarKye3IW6fPRJEATIZDKYmJggOzsbubm50NXVhVwuh4WFhVhHoVDAyMgIf//9N0xMTOpgr59S8vKjPYLwJBxl3c+Bwa3/QbP00fNXIiKieokBqZaYm5sDgBiS6jNBEHD79m2YmJigsPCfR2o8ePAAubm5So8ZKXfnzh3o6enhxo0bYlleXh4yMjJQVFSkchsPHjxATk4OFApFpc86y8nJwaNHj8RRuDqTc7dm2iktgsGt/8H8yqaaaY+IiNSCAamWyGQyWFhYwNTUtNLTVPVFVlYWPvroI2zevBkdOnQQy7dt24aTJ09i69atFdYZNGgQQkJC8M4774hlmzZtQnh4OI4dOyaWHTx4ELNmzcKjR49gYmKC8PBwtG7dWmU/bt26hQkTJuDTTz9F9+7da3APq2DVsJdvQxDQ+PE9jhwREb0CGJBqmaamZr1/MnyjRo3EB8A+fefvBw8eIDMzU+XdwG/duoXi4mKlZY8fP8adO3eUyt588038+uuvyM7OxurVq+Hr64uEhIQKp+1SU1MxYMAA9OzZE35+fjW9i8/HO14TEdFTOEmbYGxsDE1NTWRmZiqVZ2ZmiqcKpczNzatUX0dHB61bt4a7uzvWrl2LRo0aYe3atUp10tLS0KtXL3Tv3h2RkZE1sEdEREQvhwGJoKWlBWdnZ8TGxoplZWVliI2NhYeHh8p1PDw8lOoDQExMTKX1n2736XlOqamp6NmzJ5ydnbFu3boGfXk/ERG9OniKjQAAAQEBGDNmDFxcXODq6oqwsDDk5+fD398fAODn5wcrKyuEhIQAAKZPnw5PT0+EhoZiwIABiIqKQmJiojgClJ+fjyVLlmDQoEGwsLBAdnY2wsPDkZqaimHDnsz3KQ9HNjY2WL58Oe7e/WeidGUjV0RERHVBJjS0uxnWE1V9GvCLUNdTuvNO7ULeie0ozf8bWqYtYej1IeSWbQEAGZvmopG+GYwHzBTr5186ipwjP6MkNxONm1miWU9/NGnVDQAglBTh7q5lKEq7jNJHedBsogctc3vodx8OuUUbAMDDc7/j3t4wlX2xmbO7dndWgk/qVoNaelL360Jdx4nXFY8RalBLx4iq/vzmCBKJ9JwHQs95oMpl5qOWVijTcXgDOg5vqKwva6QF03c/e+b2dB29oOvoVf2OEhER1TJO+CAiIiKSYEAiIiIikmBAIiIiIpJgQCIiIiKSYEAiIiIikmBAIiIiIpJgQCIiIiKSYEAiIiIikmBAIiIiIpJgQCIiIiKSYEAiIiIikmBAIiIiIpJgQCIiIiKSqBcBKTw8HLa2tlAoFHBzc8OJEyeeWT86OhoODg5QKBRwdHTE3r17lZYHBwfDwcEBOjo6aNasGby8vJCQkKBU5/79+xg9ejT09PRgYGCA8ePH4+HDhzW+b0RERNTwqD0gbdmyBQEBAQgKCkJSUhI6d+4Mb29vZGVlqawfFxeHkSNHYvz48UhOToaPjw98fHxw/vx5sU6bNm2watUqnDt3DkePHoWtrS369u2Lu3fvinVGjx6NP//8EzExMdi9ezcOHz6MiRMn1vr+EhERUf0nEwRBUGcH3Nzc0K1bN6xatQoAUFZWBmtra0ydOhVz586tUH/48OHIz8/H7t27xTJ3d3c4OTkhIiJC5Tby8vKgr6+P33//Hb1798bFixfRvn17nDx5Ei4uLgCAffv24e2338adO3dgaWn53H6Xt5mbmws9Pb0X2fVK2c7dU6Pt0fPdVIxSdxdeP8G56u5Bg8bjRN3iMUINaukYUdWf32odQSoqKsKpU6fg5eUllmloaMDLywvx8fEq14mPj1eqDwDe3t6V1i8qKkJkZCT09fXRuXNnsQ0DAwMxHAGAl5cXNDQ0KpyKK1dYWIi8vDylFxEREb2a1BqQsrOzUVpaCjMzM6VyMzMzZGRkqFwnIyOjSvV3794NXV1dKBQKfPPNN4iJiYGxsbHYhqmpqVL9Ro0awdDQsNLthoSEQF9fX3xZW1tXa1+JiIio4VD7HKTa0qtXL5w+fRpxcXHo168ffH19K53XVBWBgYHIzc0VX7dv367B3hIREVF9otaAZGxsDE1NTWRmZiqVZ2ZmwtzcXOU65ubmVaqvo6OD1q1bw93dHWvXrkWjRo2wdu1asQ1pWCopKcH9+/cr3a5cLoeenp7Si4iIiF5Nag1IWlpacHZ2RmxsrFhWVlaG2NhYeHh4qFzHw8NDqT4AxMTEVFr/6XYLCwvFNnJycnDq1Clx+YEDB1BWVgY3N7cX3R0iIiJ6RTRSdwcCAgIwZswYuLi4wNXVFWFhYcjPz4e/vz8AwM/PD1ZWVggJCQEATJ8+HZ6enggNDcWAAQMQFRWFxMREREZGAgDy8/OxZMkSDBo0CBYWFsjOzkZ4eDhSU1MxbNgwAEC7du3Qr18/TJgwARERESguLsaUKVMwYsSIKl3BRkRERK82tQek4cOH4+7du1i4cCEyMjLg5OSEffv2iROxU1JSoKHxz0BX9+7dsWnTJsyfPx/z5s2Dvb09du7ciY4dOwIANDU1cenSJWzYsAHZ2dkwMjJCt27dcOTIEXTo0EFsZ+PGjZgyZQp69+4NDQ0NDB06FN9++23d7jwRERHVS2q/D1JDxfsgvVp4jxM14H2QXgqPE3WLxwg1eJ3vg0RERERUHzEgEREREUkwIBERERFJMCARERERSTAgEREREUkwIBERERFJMCARERERSTAgEREREUkwIBERERFJMCARERERSTAgEREREUkwIBERERFJMCARERERSTAgEREREUkwIBERERFJMCARERERSTAgEREREUkwIBERERFJMCARERERSTAgEREREUkwIBERERFJMCARERERSTAgEREREUkwIBERERFJMCARERERSTAgEREREUkwIBERERFJMCARERERSTAgEREREUkwIBERERFJMCARERERSTAgEREREUkwIBERERFJMCARERERSTAgEREREUkwIBERERFJMCARERERSdSLgBQeHg5bW1soFAq4ubnhxIkTz6wfHR0NBwcHKBQKODo6Yu/eveKy4uJizJkzB46OjtDR0YGlpSX8/PyQlpam1IatrS1kMpnSa+nSpbWyf0RERNSwqD0gbdmyBQEBAQgKCkJSUhI6d+4Mb29vZGVlqawfFxeHkSNHYvz48UhOToaPjw98fHxw/vx5AEBBQQGSkpKwYMECJCUlYfv27bh8+TIGDRpUoa3PP/8c6enp4mvq1Km1uq9ERETUMMgEQRDU2QE3Nzd069YNq1atAgCUlZXB2toaU6dOxdy5cyvUHz58OPLz87F7926xzN3dHU5OToiIiFC5jZMnT8LV1RW3bt1CixYtADwZQZoxYwZmzJjxQv3Oy8uDvr4+cnNzoaen90JtVMZ27p4abY+e76ZilLq78PoJzlV3Dxo0HifqFo8RalBLx4iq/vxW6whSUVERTp06BS8vL7FMQ0MDXl5eiI+PV7lOfHy8Un0A8Pb2rrQ+AOTm5kImk8HAwECpfOnSpTAyMkKXLl2wbNkylJSUvPjOEBER0SujkTo3np2djdLSUpiZmSmVm5mZ4dKlSyrXycjIUFk/IyNDZf3Hjx9jzpw5GDlypFJSnDZtGrp27QpDQ0PExcUhMDAQ6enpWLFihcp2CgsLUVhYKL7Py8ur0j4SERFRw6PWgFTbiouL4evrC0EQ8MMPPygtCwgIEP/eqVMnaGlp4cMPP0RISAjkcnmFtkJCQrBo0aJa7zMRERGpn1pPsRkbG0NTUxOZmZlK5ZmZmTA3N1e5jrm5eZXql4ejW7duISYm5rnzhNzc3FBSUoKbN2+qXB4YGIjc3Fzxdfv27efsHRERETVUag1IWlpacHZ2RmxsrFhWVlaG2NhYeHh4qFzHw8NDqT4AxMTEKNUvD0dXrlzB77//DiMjo+f25fTp09DQ0ICpqanK5XK5HHp6ekovIiIiejWp/RRbQEAAxowZAxcXF7i6uiIsLAz5+fnw9/cHAPj5+cHKygohISEAgOnTp8PT0xOhoaEYMGAAoqKikJiYiMjISABPwtF7772HpKQk7N69G6WlpeL8JENDQ2hpaSE+Ph4JCQno1asXmjZtivj4eMycORP//ve/0axZM/V8EERERFRvqD0gDR8+HHfv3sXChQuRkZEBJycn7Nu3T5yInZKSAg2Nfwa6unfvjk2bNmH+/PmYN28e7O3tsXPnTnTs2BEAkJqail9//RUA4OTkpLStgwcPomfPnpDL5YiKikJwcDAKCwthZ2eHmTNnKs1LIiIioteX2u+D1FDxPkivFt7jRA14H6SXwuNE3eIxQg1e5/sgEREREdVHDEhEREREEgxIRERERBIMSEREREQSDEhEREREEgxIRERERBIMSEREREQSDEhEREREEgxIRERERBIMSEREREQSDEhEREREEgxIRERERBIMSEREREQSDEhEREREEgxIRERERBIMSEREREQSDEhEREREEgxIRERERBIMSEREREQSDEhEREREEgxIRERERBIMSEREREQSDEhEREREEgxIRERERBIMSEREREQSDEhEREREEgxIRERERBIMSEREREQSDEhEREREEgxIRERERBIMSEREREQSDEhEREREEgxIRERERBIMSEREREQSDEhEREREEgxIRERERBIMSEREREQSDEhEREREEvUiIIWHh8PW1hYKhQJubm44ceLEM+tHR0fDwcEBCoUCjo6O2Lt3r7isuLgYc+bMgaOjI3R0dGBpaQk/Pz+kpaUptXH//n2MHj0aenp6MDAwwPjx4/Hw4cNa2T8iIiJqWNQekLZs2YKAgAAEBQUhKSkJnTt3hre3N7KyslTWj4uLw8iRIzF+/HgkJyfDx8cHPj4+OH/+PACgoKAASUlJWLBgAZKSkrB9+3ZcvnwZgwYNUmpn9OjR+PPPPxETE4Pdu3fj8OHDmDhxYq3vLxEREdV/MkEQhKpWPnHiBJydnaGpqalyeWFhIf773//C19e3yh1wc3NDt27dsGrVKgBAWVkZrK2tMXXqVMydO7dC/eHDhyM/Px+7d+8Wy9zd3eHk5ISIiAiV2zh58iRcXV1x69YttGjRAhcvXkT79u1x8uRJuLi4AAD27duHt99+G3fu3IGlpeVz+52Xlwd9fX3k5uZCT0+vyvtbFbZz99Roe/R8NxWj1N2F109wrrp70KDxOFG3eIxQg1o6RlT153e1RpA8PDxw79498b2enh6uX78uvs/JycHIkSOr3F5RURFOnToFLy+vfzqkoQEvLy/Ex8erXCc+Pl6pPgB4e3tXWh8AcnNzIZPJYGBgILZhYGAghiMA8PLygoaGBhISElS2UVhYiLy8PKUXERERvZqqFZCkg02qBp+qMSCF7OxslJaWwszMTKnczMwMGRkZKtfJyMioVv3Hjx9jzpw5GDlypJgUMzIyYGpqqlSvUaNGMDQ0rLSdkJAQ6Ovriy9ra+sq7SMRERE1PDU+B0kmk9V0ky+suLgYvr6+EAQBP/zww0u1FRgYiNzcXPF1+/btGuolERER1TeN1LlxY2NjaGpqIjMzU6k8MzMT5ubmKtcxNzevUv3ycHTr1i0cOHBA6Tyjubl5hUngJSUluH//fqXblcvlkMvlVd43IiIiariqPYJ04cIFnD17FmfPnoUgCLh06ZL4/s8//6xWW1paWnB2dkZsbKxYVlZWhtjYWHh4eKhcx8PDQ6k+AMTExCjVLw9HV65cwe+//w4jI6MKbeTk5ODUqVNi2YEDB1BWVgY3N7dq7QMRERG9eqo9gtS7d2+leUbvvPMOgCen1gRBqPYptoCAAIwZMwYuLi5wdXVFWFgY8vPz4e/vDwDw8/ODlZUVQkJCAADTp0+Hp6cnQkNDMWDAAERFRSExMRGRkZEAnoSj9957D0lJSdi9ezdKS0vFeUWGhobQ0tJCu3bt0K9fP0yYMAEREREoLi7GlClTMGLEiCpdwUZERESvtmoFpBs3btR4B4YPH467d+9i4cKFyMjIgJOTE/bt2ydOxE5JSYGGxj8DXd27d8emTZswf/58zJs3D/b29ti5cyc6duwIAEhNTcWvv/4KAHByclLa1sGDB9GzZ08AwMaNGzFlyhT07t0bGhoaGDp0KL799tsa3z8iIiJqeKp1H6SqOH/+vBhWXmW8D9Krhfc4UQPeB+ml8DhRt3iMUIOGdB+kyjx48ACRkZFwdXVF586da6JJIiIiIrV5qYB0+PBhjBkzBhYWFli+fDneeustHD9+vKb6RkRERKQW1Z6knZGRgfXr12Pt2rXIy8uDr68vCgsLsXPnTrRv3742+khERERUp6o1gjRw4EC0bdsWZ8+eRVhYGNLS0vDdd9/VVt+IiIiI1KJaI0j/+9//MG3aNEyaNAn29va11SciIiIitarWCNLRo0fx4MEDODs7w83NDatWrUJ2dnZt9Y2IiIhILaoVkNzd3bF69Wqkp6fjww8/RFRUFCwtLVFWVoaYmBg8ePCgtvpJREREVGde6Co2HR0djBs3DkePHsW5c+cwa9YsLF26FKamphg0aFBN95GIiIioTr30fZDatm2Lr7/+Gnfu3EFUVFS1HzVCREREVN9Ua5L2uHHjnltH+mBYIiIiooamWgFp/fr1sLGxQZcuXVDZE0o4gkREREQNXbUC0qRJk7B582bcuHED/v7++Pe//w1DQ8Pa6hsRERGRWlRrDlJ4eDjS09Px6aefYteuXbC2toavry9+++23SkeUiIiIiBqaak/SlsvlGDlyJGJiYnDhwgV06NABH3/8MWxtbfHw4cPa6CMRERFRnXqpq9g0NDQgk8kgCAJKS0trqk9EREREalXtgFRYWIjNmzejT58+aNOmDc6dO4dVq1YhJSUFurq6tdFHIiIiojpVrUnaH3/8MaKiomBtbY1x48Zh8+bNMDY2rq2+EREREalFtQJSREQEWrRogZYtW+KPP/7AH3/8obLe9u3ba6RzREREROpQrYDk5+fH+xwRERHRK6/aN4okIiIietW99LPYiIiIiF41DEhEREREEgxIRERERBIMSEREREQSDEhEREREEgxIRERERBIMSEREREQSDEhEREREEgxIRERERBIMSEREREQSDEhEREREEgxIRERERBIMSEREREQSDEhEREREEgxIRERERBIMSEREREQSDEhEREREEgxIRERERBJqD0jh4eGwtbWFQqGAm5sbTpw48cz60dHRcHBwgEKhgKOjI/bu3au0fPv27ejbty+MjIwgk8lw+vTpCm307NkTMplM6fXRRx/V5G4RERFRA6bWgLRlyxYEBAQgKCgISUlJ6Ny5M7y9vZGVlaWyflxcHEaOHInx48cjOTkZPj4+8PHxwfnz58U6+fn5eOONN/DVV189c9sTJkxAenq6+Pr6669rdN+IiIio4VJrQFqxYgUmTJgAf39/tG/fHhEREdDW1saPP/6osv7KlSvRr18/zJ49G+3atcPixYvRtWtXrFq1Sqzz/vvvY+HChfDy8nrmtrW1tWFubi6+9PT0anTfiIiIqOFSW0AqKirCqVOnlIKMhoYGvLy8EB8fr3Kd+Pj4CsHH29u70vrPsnHjRhgbG6Njx44IDAxEQUHBM+sXFhYiLy9P6UVERESvpkbq2nB2djZKS0thZmamVG5mZoZLly6pXCcjI0Nl/YyMjGpte9SoUbCxsYGlpSXOnj2LOXPm4PLly9i+fXul64SEhGDRokXV2g4RERE1TGoLSOo0ceJE8e+Ojo6wsLBA7969ce3aNbRq1UrlOoGBgQgICBDf5+Xlwdrautb7SkRERHVPbQHJ2NgYmpqayMzMVCrPzMyEubm5ynXMzc2rVb+q3NzcAABXr16tNCDJ5XLI5fKX2g4RERE1DGqbg6SlpQVnZ2fExsaKZWVlZYiNjYWHh4fKdTw8PJTqA0BMTEyl9auq/FYAFhYWL9UOERERvRrUeootICAAY8aMgYuLC1xdXREWFob8/Hz4+/sDAPz8/GBlZYWQkBAAwPTp0+Hp6YnQ0FAMGDAAUVFRSExMRGRkpNjm/fv3kZKSgrS0NADA5cuXAUC8Wu3atWvYtGkT3n77bRgZGeHs2bOYOXMm3nzzTXTq1KmOPwEiIiKqj9QakIYPH467d+9i4cKFyMjIgJOTE/bt2ydOxE5JSYGGxj+DXN27d8emTZswf/58zJs3D/b29ti5cyc6duwo1vn111/FgAUAI0aMAAAEBQUhODgYWlpa+P3338UwZm1tjaFDh2L+/Pl1tNdERERU38kEQRDU3YmGKC8vD/r6+sjNza3xeyjZzt1To+3R891UjFJ3F14/wbnq7kGDxuNE3eIxQg1q6RhR1Z/fan/UCBEREVF9w4BEREREJMGARERERCTBgEREREQkwYBEREREJMGARERERCTBgEREREQkwYBEREREJMGARERERCTBgEREREQkwYBEREREJMGARERERCTBgEREREQkwYBEREREJMGARERERCTBgEREREQkwYBEREREJMGARERERCTBgEREREQkwYBEREREJMGARERERCTBgEREREQkwYBEREREJMGARERERCTBgEREREQkwYBEREREJMGARERERCTBgEREREQkwYBEREREJMGARERERCTBgEREREQkwYBEREREJMGARERERCTBgEREREQkwYBEREREJMGARERERCTBgEREREQkofaAFB4eDltbWygUCri5ueHEiRPPrB8dHQ0HBwcoFAo4Ojpi7969Ssu3b9+Ovn37wsjICDKZDKdPn67QxuPHjzF58mQYGRlBV1cXQ4cORWZmZk3uFhERETVgag1IW7ZsQUBAAIKCgpCUlITOnTvD29sbWVlZKuvHxcVh5MiRGD9+PJKTk+Hj4wMfHx+cP39erJOfn4833ngDX331VaXbnTlzJnbt2oXo6Gj88ccfSEtLw5AhQ2p8/4iIiKhhkgmCIKhr425ubujWrRtWrVoFACgrK4O1tTWmTp2KuXPnVqg/fPhw5OfnY/fu3WKZu7s7nJycEBERoVT35s2bsLOzQ3JyMpycnMTy3NxcmJiYYNOmTXjvvfcAAJcuXUK7du0QHx8Pd3f3KvU9Ly8P+vr6yM3NhZ6eXnV3/Zls5+6p0fbo+W4qRqm7C6+f4Fx196BB43GibvEYoQa1dIyo6s9vtY0gFRUV4dSpU/Dy8vqnMxoa8PLyQnx8vMp14uPjleoDgLe3d6X1VTl16hSKi4uV2nFwcECLFi2e2U5hYSHy8vKUXkRERPRqUltAys7ORmlpKczMzJTKzczMkJGRoXKdjIyMatWvrA0tLS0YGBhUq52QkBDo6+uLL2tr6ypvk4iIiBoWtU/SbigCAwORm5srvm7fvq3uLhEREVEtaaSuDRsbG0NTU7PC1WOZmZkwNzdXuY65uXm16lfWRlFREXJycpRGkZ7Xjlwuh1wur/J2iIiIqOFS2wiSlpYWnJ2dERsbK5aVlZUhNjYWHh4eKtfx8PBQqg8AMTExldZXxdnZGY0bN1Zq5/Lly0hJSalWO0RERPTqUtsIEgAEBARgzJgxcHFxgaurK8LCwpCfnw9/f38AgJ+fH6ysrBASEgIAmD59Ojw9PREaGooBAwYgKioKiYmJiIyMFNu8f/8+UlJSkJaWBuBJ+AGejByZm5tDX18f48ePR0BAAAwNDaGnp4epU6fCw8OjylewERER0atNrQFp+PDhuHv3LhYuXIiMjAw4OTlh37594kTslJQUaGj8M8jVvXt3bNq0CfPnz8e8efNgb2+PnTt3omPHjmKdX3/9VQxYADBixAgAQFBQEIKDgwEA33zzDTQ0NDB06FAUFhbC29sb33//fR3sMRERETUEar0PUkPG+yC9WniPEzXgfZBeCo8TdYvHCDV4Xe+DRERERFRfMSARERERSTAgEREREUkwIBERERFJMCARERERSTAgEREREUkwIBERERFJMCARERERSTAgEREREUkwIBERERFJMCARERERSTAgEREREUkwIBERERFJMCARERERSTAgEREREUkwIBERERFJMCARERERSTAgEREREUkwIBERERFJMCARERERSTAgEREREUkwIBERERFJMCARERERSTAgEREREUkwIBERERFJMCARERERSTAgEREREUkwIBERERFJMCARERERSTAgEREREUkwIBERERFJMCARERERSTAgEREREUkwIBERERFJMCARERERSTAgEREREUkwIBERERFJMCARERERSdSLgBQeHg5bW1soFAq4ubnhxIkTz6wfHR0NBwcHKBQKODo6Yu/evUrLBUHAwoULYWFhgSZNmsDLywtXrlxRqmNrawuZTKb0Wrp0aY3vGxERETU8ag9IW7ZsQUBAAIKCgpCUlITOnTvD29sbWVlZKuvHxcVh5MiRGD9+PJKTk+Hj4wMfHx+cP39erPP111/j22+/RUREBBISEqCjowNvb288fvxYqa3PP/8c6enp4mvq1Km1uq9ERETUMKg9IK1YsQITJkyAv78/2rdvj4iICGhra+PHH39UWX/lypXo168fZs+ejXbt2mHx4sXo2rUrVq1aBeDJ6FFYWBjmz5+PwYMHo1OnTvjpp5+QlpaGnTt3KrXVtGlTmJubiy8dHZ3a3l0iIiJqANQakIqKinDq1Cl4eXmJZRoaGvDy8kJ8fLzKdeLj45XqA4C3t7dY/8aNG8jIyFCqo6+vDzc3twptLl26FEZGRujSpQuWLVuGkpKSSvtaWFiIvLw8pRcRERG9mhqpc+PZ2dkoLS2FmZmZUrmZmRkuXbqkcp2MjAyV9TMyMsTl5WWV1QGAadOmoWvXrjA0NERcXBwCAwORnp6OFStWqNxuSEgIFi1aVL0dJCIiogZJrQFJnQICAsS/d+rUCVpaWvjwww8REhICuVxeoX5gYKDSOnl5ebC2tq6TvhIREVHdUuspNmNjY2hqaiIzM1OpPDMzE+bm5irXMTc3f2b98j+r0yYAuLm5oaSkBDdv3lS5XC6XQ09PT+lFRERErya1BiQtLS04OzsjNjZWLCsrK0NsbCw8PDxUruPh4aFUHwBiYmLE+nZ2djA3N1eqk5eXh4SEhErbBIDTp09DQ0MDpqamL7NLRERE9ApQ+ym2gIAAjBkzBi4uLnB1dUVYWBjy8/Ph7+8PAPDz84OVlRVCQkIAANOnT4enpydCQ0MxYMAAREVFITExEZGRkQAAmUyGGTNm4IsvvoC9vT3s7OywYMECWFpawsfHB8CTid4JCQno1asXmjZtivj4eMycORP//ve/0axZM7V8DkRERFR/qD0gDR8+HHfv3sXChQuRkZEBJycn7Nu3T5xknZKSAg2Nfwa6unfvjk2bNmH+/PmYN28e7O3tsXPnTnTs2FGs8+mnnyI/Px8TJ05ETk4O3njjDezbtw8KhQLAk9NlUVFRCA4ORmFhIezs7DBz5kylOUZERET0+pIJgiCouxMNUV5eHvT19ZGbm1vj85Fs5+6p0fbo+W4qRqm7C6+f4Fx196BB43GibvEYoQa1dIyo6s9vtd8okoiIiKi+YUAiIiIikmBAIiIiIpJgQCIiIiKSYEAiIiIikmBAIiIiIpJgQCIiIiKSYEAiIiIikmBAIiIiIpJgQCIiIiKSYEAiIiIikmBAIiIiIpJgQCIiIiKSYEAiIiIikmBAIiIiIpJgQCIiIiKSYEAiIiIikmBAIiIiIpJgQCIiIiKSYEAiIiIikmBAIiIiIpJgQCIiIiKSYEAiIiIikmBAIiIiIpJgQCIiIiKSYEAiIiIikmBAIiIiIpJgQCIiIiKSYEAiIiIikmBAIiIiIpJgQCIiIiKSYEAiIiIikmBAIiIiIpJgQCIiIiKSYEAiIiIikmBAIiIiIpJgQCIiIiKSqBcBKTw8HLa2tlAoFHBzc8OJEyeeWT86OhoODg5QKBRwdHTE3r17lZYLgoCFCxfCwsICTZo0gZeXF65cuaJU5/79+xg9ejT09PRgYGCA8ePH4+HDhzW+b0RERNTwqD0gbdmyBQEBAQgKCkJSUhI6d+4Mb29vZGVlqawfFxeHkSNHYvz48UhOToaPjw98fHxw/vx5sc7XX3+Nb7/9FhEREUhISICOjg68vb3x+PFjsc7o0aPx559/IiYmBrt378bhw4cxceLEWt9fIiIiqv9kgiAI6uyAm5sbunXrhlWrVgEAysrKYG1tjalTp2Lu3LkV6g8fPhz5+fnYvXu3WObu7g4nJydERERAEARYWlpi1qxZ+OSTTwAAubm5MDMzw/r16zFixAhcvHgR7du3x8mTJ+Hi4gIA2LdvH95++23cuXMHlpaWz+13Xl4e9PX1kZubCz09vZr4KES2c/fUaHv0fDcVo9TdhddPcK66e9Cg8ThRt3iMUINaOkZU9ee3WkeQioqKcOrUKXh5eYllGhoa8PLyQnx8vMp14uPjleoDgLe3t1j/xo0byMjIUKqjr68PNzc3sU58fDwMDAzEcAQAXl5e0NDQQEJCQo3tHxERETVMjdS58ezsbJSWlsLMzEyp3MzMDJcuXVK5TkZGhsr6GRkZ4vLysmfVMTU1VVreqFEjGBoainWkCgsLUVhYKL7PzX2SbPPy8p65jy+irLCgxtukZ8uTqXUg9fVUC/93Xic8TtQtHiPUoJaOEeU/t593Ak2tAakhCQkJwaJFiyqUW1tbq6E3VNP01d2B19FSfurUcPBfqxrU8jHiwYMH0NevfBtqDUjGxsbQ1NREZmamUnlmZibMzc1VrmNubv7M+uV/ZmZmwsLCQqmOk5OTWEc6CbykpAT379+vdLuBgYEICAgQ35eVleH+/fswMjKCTCarwt5SfZWXlwdra2vcvn27xueTEVHDx2PEq0UQBDx48OC5843VGpC0tLTg7OyM2NhY+Pj4AHgSPGJjYzFlyhSV63h4eCA2NhYzZswQy2JiYuDh4QEAsLOzg7m5OWJjY8VAlJeXh4SEBEyaNElsIycnB6dOnYKzszMA4MCBAygrK4Obm5vK7crlcsjlcqUyAwODF9xzqo/09PR48COiSvEY8ep41shRObWfYgsICMCYMWPg4uICV1dXhIWFIT8/H/7+/gAAPz8/WFlZISQkBAAwffp0eHp6IjQ0FAMGDEBUVBQSExMRGRkJAJDJZJgxYwa++OIL2Nvbw87ODgsWLIClpaUYwtq1a4d+/fphwoQJiIiIQHFxMaZMmYIRI0ZU6Qo2IiIierWpPSANHz4cd+/excKFC5GRkQEnJyfs27dPnGSdkpICDY1/Lrbr3r07Nm3ahPnz52PevHmwt7fHzp070bFjR7HOp59+ivz8fEycOBE5OTl44403sG/fPigUCrHOxo0bMWXKFPTu3RsaGhoYOnQovv3227rbcSIiIqq31H4fJCJ1KywsREhICAIDAyucRiUi4jHi9cSARERERCSh9keNEBEREdU3DEhEREREEgxIRERERBIMSETVdPPmTchkMpw+fVrdXSGiZ1i/fn29vF/doUOHIJPJkJOTo+6u0DMwIFG9JJPJnvkKDg6ule0GBwc/d9vW1tZIT09XurUEEdWe27dvY9y4cbC0tISWlhZsbGwwffp03Lt3T6xja2uLsLAw9XXy/xs7duwzjx+2trbo3r070tPTq3SzQlIfBiSql9LT08VXWFgY9PT0lMo++eSTWtnuJ598orSd5s2b4/PPP1cq09TUhLm5ORo1evHbiBUVFdVgr4leXdevX4eLiwuuXLmCzZs34+rVq4iIiEBsbCw8PDxw//79Ou9TcXFxpctWrlypdLwAgHXr1onvT548CS0tLZibm7/UY6p4DKl9DEhUL5mbm4svfX19yGQy8b2pqSlWrFiB5s2bQy6XizcXLVd+CiwqKgrdu3eHQqFAx44d8ccffzx3u7q6ukrb1tTURNOmTZXKVJ1iO3/+PPr37w9dXV2YmZnh/fffR3Z2tri8Z8+emDJlCmbMmAFjY2N4e3vX6OdF9KqaPHkytLS0sH//fnh6eqJFixbo378/fv/9d6SmpuKzzz5Dz549cevWLcycOVMcqXnab7/9hnbt2kFXVxf9+vUTg0u5NWvWoF27dlAoFHBwcMD3338vLiv//75lyxZ4enpCoVBg48aNlfZXX19f6XgBPHksVfl7ExMTlafYjh49in/9619o0qQJrK2tMW3aNOTn54vLbW1tsXjxYvj5+UFPTw8TJ058mY+VqoABiRqclStXIjQ0FMuXL8fZs2fh7e2NQYMG4cqVK0r1Zs+ejVmzZiE5ORkeHh4YOHCg0pB8TcnJycFbb72FLl26IDExEfv27UNmZiZ8fX2V6m3YsAFaWlo4duwYIiIiarwfRK+a+/fv47fffsPHH3+MJk2aKC0zNzfH6NGjsWXLFvzyyy8VRnvLFRQUYPny5fj5559x+PBhpKSkKI1Ab9y4EQsXLsSSJUtw8eJFfPnll1iwYAE2bNigtL25c+di+vTpuHjxYo3/gnPt2jX069cPQ4cOxdmzZ7FlyxYcPXq0wjNJly9fjs6dOyM5ORkLFiyo0T6QCgJRPbdu3TpBX19ffG9paSksWbJEqU63bt2Ejz/+WBAEQbhx44YAQFi6dKm4vLi4WGjevLnw1VdfVWvbNjY2wjfffKNUVt5+cnKyIAiCsHjxYqFv375KdW7fvi0AEC5fviwIgiB4enoKXbp0qda2iV53x48fFwAIO3bsULl8xYoVAgAhMzNT5f/VdevWCQCEq1evimXh4eGCmZmZ+L5Vq1bCpk2blNZbvHix4OHhIQjCP//fw8LCXmgfVPX/4MGDAgDh77//FgRBEMaPHy9MnDhRqc6RI0cEDQ0N4dGjR4IgPDkW+fj4vFAf6MWo/VlsRNWRl5eHtLQ09OjRQ6m8R48eOHPmjFKZh4eH+PdGjRrBxcUFFy9erPE+nTlzBgcPHoSurm6FZdeuXUObNm0AAM7OzjW+baLXgfASD3zQ1tZGq1atxPcWFhbIysoCAOTn5+PatWsYP348JkyYINYpKSmpMIHaxcXlhfvwPGfOnMHZs2eVTt0JgoCysjLcuHED7dq1q/U+UEUMSEQv6eHDhxg4cCC++uqrCsssLCzEv+vo6NRlt4gavNatW0Mmk+HixYt49913Kyy/ePEimjVrBhMTk0rbaNy4sdJ7mUwmBq6HDx8CAFavXg03Nzelepqamkrva/P/78OHD/Hhhx9i2rRpFZa1aNGiTvpAFTEgUYOip6cHS0tLHDt2DJ6enmL5sWPH4OrqqlT3+PHjePPNNwE8+Y3w1KlTFc7p14SuXbvil19+ga2t7Utd2UZEyoyMjNCnTx98//33mDlzptI8pIyMDGzcuBF+fn6QyWTQ0tJCaWlptdo3MzODpaUlrl+/jtGjR9d096usa9euuHDhAlq3bq22PlBFnKRNDc7s2bPx1VdfYcuWLbh8+TLmzp2L06dPY/r06Ur1wsPDsWPHDly6dAmTJ0/G33//jXHjxtV4fyZPnoz79+9j5MiROHnyJK5du4bffvsN/v7+1T5gE5GyVatWobCwEN7e3jh8+DBu376Nffv2oU+fPrCyssKSJUsAPLnK6/Dhw0hNTVW6gvR5Fi1ahJCQEHz77bf466+/cO7cOaxbtw4rVqyorV2qYM6cOYiLi8OUKVNw+vRpXLlyBf/9739r5Rc6qjoGJGpwpk2bhoCAAMyaNQuOjo7Yt28ffv31V9jb2yvVW7p0KZYuXYrOnTvj6NGj+PXXX2FsbFzj/Skf0SotLUXfvn3h6OiIGTNmwMDAABoa/C9G9DLs7e2RmJiIli1bwtfXF61atcLEiRPRq1cvxMfHw9DQEADw+eef4+bNm2jVqtUzT7lJffDBB1izZg3WrVsHR0dHeHp6Yv369bCzs6utXaqgU6dO+OOPP/DXX3/hX//6F7p06YKFCxfC0tKyzvpAFcmEl5n9RlQP3bx5E3Z2dkhOToaTk5O6u0NERA0Qf70lIiIikmBAotfORx99BF1dXZWvjz76SN3dI6J6rvyu+apeX375pbq7RzWEp9jotZOVlYW8vDyVy/T09GBqalrHPSKihiQ1NRWPHj1SuczQ0FCcF0UNGwMSERERkQRPsRERERFJMCARERERSTAgEREREUkwIBERERFJMCAR0SsjIyMD06dPR+vWraFQKGBmZoYePXrghx9+QEFBgbq7R0QNCJ+sSUSvhOvXr6NHjx4wMDDAl19+CUdHR8jlcpw7dw6RkZGwsrLCoEGDKqxXXFxc4YnvREQcQSKiV8LHH3+MRo0aITExEb6+vmjXrh1atmyJwYMHY8+ePRg4cCAAQCaT4YcffsCgQYOgo6MjPuz0hx9+QKtWraClpYW2bdvi559/Ftu+efMmZDIZTp8+LZbl5ORAJpPh0KFDAIBDhw5BJpNhz5496NSpExQKBdzd3XH+/Pk6+wyIqOYwIBFRg3fv3j3s378fkydPho6Ojso6MplM/HtwcDDeffddnDt3DuPGjcOOHTswffp0zJo1C+fPn8eHH34If39/HDx4sNp9mT17NkJDQ3Hy5EmYmJhg4MCBKC4ufuF9IyL1YEAiogbv6tWrEAQBbdu2VSo3NjYWHwExZ84csXzUqFHw9/dHy5Yt0aJFCyxfvhxjx47Fxx9/jDZt2iAgIABDhgzB8uXLq92XoKAg9OnTB46OjtiwYQMyMzOxY8eOl95HIqpbDEhE9Mo6ceIETp8+jQ4dOqCwsFAsd3FxUap38eJF9OjRQ6msR48euHjxYrW36eHhIf7d0NAQbdu2faF2iEi9OEmbiBq81q1bQyaT4fLly0rlLVu2BAA0adJEqbyy03CV0dB48rvk009m4mkzolcbR5CIqMEzMjJCnz59sGrVKuTn51d7/Xbt2uHYsWNKZceOHUP79u0BACYmJgCA9PR0cfnTE7afdvz4cfHvf//9N/766y+0a9eu2n0iIvXiCBIRvRK+//579OjRAy4uLggODkanTp2goaGBkydP4tKlS3B2dq503dmzZ8PX1xddunSBl5cXdu3ahe3bt+P3338H8GQEyt3dHUuXLoWdnR2ysrIwf/58lW19/vnnMDIygpmZGT777DMYGxvDx8enNnaZiGqTQET0ikhLSxOmTJki2NnZCY0bNxZ0dXUFV1dXYdmyZUJ+fr4gCIIAQNixY0eFdb///nuhZcuWQuPGjYU2bdoIP/30k9LyCxcuCB4eHkKTJk0EJycnYf/+/QIA4eDBg4IgCMLBgwcFAMKuXbuEDh06CFpaWoKrq6tw5syZ2t5tIqoFMkF46qQ6ERG9kEOHDqFXr174+++/YWBgoO7uENFL4hwkIiIiIgkGJCIiIiIJnmIjIiIikuAIEhEREZEEAxIRERGRBAMSERERkQQDEhEREZEEAxIRERGRBAMSERERkQQDEhEREZEEAxIRERGRBAMSERERkcT/A6P8AJNdLfzSAAAAAElFTkSuQmCC\n"},"metadata":{}}]}]}