[Autogluon] df_train1, df_test

Author

김보람

Published

May 3, 2024

imports

import pandas as pd
import numpy as np
import sklearn
import pickle 
import time 
import datetime
import warnings
from autogluon.tabular import TabularDataset, TabularPredictor

warnings.filterwarnings('ignore')
df_train1 = pd.read_csv('~/Dropbox/Data/df_train1.csv')
df_train2 = pd.read_csv('~/Dropbox/Data/df_train2.csv')
df_train3 = pd.read_csv('~/Dropbox/Data/df_train3.csv')
df_train4 = pd.read_csv('~/Dropbox/Data/df_train4.csv')
df_train5 = pd.read_csv('~/Dropbox/Data/df_train5.csv')
df_train6 = pd.read_csv('~/Dropbox/Data/df_train6.csv')
df_train7 = pd.read_csv('~/Dropbox/Data/df_train7.csv')
df_train8 = pd.read_csv('~/Dropbox/Data/df_train8.csv')
df_test = pd.read_csv('~/Dropbox/Data/df_test.csv')
(df_train1.shape, df_train1.is_fraud.mean()), (df_test.shape, df_test.is_fraud.mean())
(((734003, 22), 0.005728859418830713), ((314572, 22), 0.005725239372862174))
_df1 = pd.concat([df_train1, df_test])
_df2 = pd.concat([df_train2, df_test])
_df3 = pd.concat([df_train3, df_test])
_df4 = pd.concat([df_train4, df_test])
_df5 = pd.concat([df_train5, df_test])
_df6 = pd.concat([df_train6, df_test])
_df7 = pd.concat([df_train7, df_test])
_df8 = pd.concat([df_train8, df_test])
_df1_mean = _df1.is_fraud.mean()
_df2_mean = _df2.is_fraud.mean()
_df3_mean = _df3.is_fraud.mean()
_df4_mean = _df4.is_fraud.mean()
_df5_mean = _df5.is_fraud.mean()
_df6_mean = _df6.is_fraud.mean()
_df7_mean = _df7.is_fraud.mean()
_df8_mean = _df8.is_fraud.mean()
def auto_amt_ver0503(df_tr, df_tst, _df_mean):
    df_tr = df_tr[["amt","is_fraud"]]
    df_tst = df_tst[["amt","is_fraud"]]     
    tr = TabularDataset(df_tr)
    tst = TabularDataset(df_tst)
    predictr = TabularPredictor(label="is_fraud", verbosity=1)
    t1 = time.time()
    predictr.fit(tr)
    t2 = time.time()
    time_diff = t2 - t1
    models = predictr._trainer.model_graph.nodes
    results = []
    for model_name in models:
    # 모델 평가
        eval_result = predictr.evaluate(tst, model=model_name)

    # 결과를 데이터프레임에 추가
        results.append({'model': model_name, 
                        'acc': eval_result['accuracy'], 
                        'pre': eval_result['precision'], 
                        'rec': eval_result['recall'], 
                        'f1': eval_result['f1'], 
                        'auc': eval_result['roc_auc']})
        
    model = []
    time_diff = []
    acc = []
    pre = []
    rec = []
    f1 = [] 
    auc = [] 
    graph_based = []
    method = [] 
    throw_rate = [] 
    train_size = []
    train_cols = []
    train_frate = []
    test_size = []
    test_frate = []
    hyper_params = [] 
    
    for result in results:
        model_name = result['model']
        model.append(model_name)
        time_diff.append(None)  # 각 모델별로 학습한 시간을 나타내고 싶은데 잘 안됨
        acc.append(result['acc']) 
        pre.append(result['pre'])
        rec.append(result['rec'])
        f1.append(result['f1'])
        auc.append(result['auc'])
        graph_based.append(False) 
        method.append('Autogluon') 
        throw_rate.append(_df_mean)
        train_size.append(len(tr))
        train_cols.append([col for col in tr.columns if col != 'is_fraud'])
        train_frate.append(tr.is_fraud.mean())
        test_size.append(len(tst))
        test_frate.append(tst.is_fraud.mean())
        hyper_params.append(None)
        
    df_results = pd.DataFrame(dict(
        model=model,
        time=time_diff,
        acc=acc,
        pre=pre,
        rec=rec,
        f1=f1,
        auc=auc,
        graph_based=graph_based,
        method=method,
        throw_rate=throw_rate,  
        train_size=train_size,
        train_cols=train_cols,
        train_frate=train_frate,
        test_size=test_size,
        test_frate=test_frate,
        hyper_params=hyper_params
    ))    
    ymdhms = datetime.datetime.fromtimestamp(time.time()).strftime('%Y%m%d-%H%M%S') 
    df_results.to_csv(f'../results2/{ymdhms}-Autogluon.csv',index=False)
    return df_results
auto_amt_ver0503(df_train1, df_test, _df1_mean)
No path specified. Models will be saved in: "AutogluonModels/ag-20240520_052634/"
AutoGluon infers your prediction problem is: 'binary' (because only two unique label-values observed).
    If 'binary' is not the correct problem_type, please manually specify the problem_type parameter during predictor init (You may specify problem_type as one of: ['binary', 'multiclass', 'regression'])
model time acc pre rec f1 auc graph_based method throw_rate train_size train_cols train_frate test_size test_frate hyper_params
0 KNeighborsUnif None 0.994218 0.489865 0.241532 0.323540 0.782174 False Autogluon 0.005728 734003 [amt] 0.005729 314572 0.005725 None
1 KNeighborsDist None 0.993582 0.403197 0.252082 0.310215 0.750937 False Autogluon 0.005728 734003 [amt] 0.005729 314572 0.005725 None
2 LightGBMXT None 0.994275 0.000000 0.000000 0.000000 0.765047 False Autogluon 0.005728 734003 [amt] 0.005729 314572 0.005725 None
3 LightGBM None 0.994275 0.000000 0.000000 0.000000 0.957828 False Autogluon 0.005728 734003 [amt] 0.005729 314572 0.005725 None
4 RandomForestGini None 0.993140 0.363844 0.264853 0.306555 0.816185 False Autogluon 0.005728 734003 [amt] 0.005729 314572 0.005725 None
5 RandomForestEntr None 0.993140 0.363844 0.264853 0.306555 0.816185 False Autogluon 0.005728 734003 [amt] 0.005729 314572 0.005725 None
6 CatBoost None 0.994275 0.000000 0.000000 0.000000 0.870449 False Autogluon 0.005728 734003 [amt] 0.005729 314572 0.005725 None
7 ExtraTreesGini None 0.993483 0.393316 0.254858 0.309299 0.840480 False Autogluon 0.005728 734003 [amt] 0.005729 314572 0.005725 None
8 ExtraTreesEntr None 0.993493 0.393966 0.253748 0.308680 0.839967 False Autogluon 0.005728 734003 [amt] 0.005729 314572 0.005725 None
9 NeuralNetFastAI None 0.994275 0.000000 0.000000 0.000000 0.943377 False Autogluon 0.005728 734003 [amt] 0.005729 314572 0.005725 None
10 XGBoost None 0.994275 0.000000 0.000000 0.000000 0.937504 False Autogluon 0.005728 734003 [amt] 0.005729 314572 0.005725 None
11 NeuralNetTorch None 0.994443 0.573003 0.115491 0.192237 0.957382 False Autogluon 0.005728 734003 [amt] 0.005729 314572 0.005725 None
12 LightGBMLarge None 0.994275 0.000000 0.000000 0.000000 0.960352 False Autogluon 0.005728 734003 [amt] 0.005729 314572 0.005725 None
13 WeightedEnsemble_L2 None 0.994408 0.535593 0.175458 0.264325 0.956493 False Autogluon 0.005728 734003 [amt] 0.005729 314572 0.005725 None 
auto_amt_ver0503(df_train2, df_test, _df2_mean)
No path specified. Models will be saved in: "AutogluonModels/ag-20240520_053531/"
AutoGluon infers your prediction problem is: 'binary' (because only two unique label-values observed).
    If 'binary' is not the correct problem_type, please manually specify the problem_type parameter during predictor init (You may specify problem_type as one of: ['binary', 'multiclass', 'regression'])
model time acc pre rec f1 auc graph_based method throw_rate train_size train_cols train_frate test_size test_frate hyper_params
0 KNeighborsUnif None 0.993753 0.440837 0.339256 0.383433 0.816442 False Autogluon 0.008171 420500 [amt] 0.01 314572 0.005725 None
1 KNeighborsDist None 0.992984 0.378443 0.350916 0.364160 0.793232 False Autogluon 0.008171 420500 [amt] 0.01 314572 0.005725 None
2 LightGBMXT None 0.994275 0.000000 0.000000 0.000000 0.765797 False Autogluon 0.008171 420500 [amt] 0.01 314572 0.005725 None
3 LightGBM None 0.994275 0.000000 0.000000 0.000000 0.957293 False Autogluon 0.008171 420500 [amt] 0.01 314572 0.005725 None
4 RandomForestGini None 0.992393 0.341880 0.355358 0.348489 0.840338 False Autogluon 0.008171 420500 [amt] 0.01 314572 0.005725 None
5 RandomForestEntr None 0.992393 0.341880 0.355358 0.348489 0.840338 False Autogluon 0.008171 420500 [amt] 0.01 314572 0.005725 None
6 CatBoost None 0.994253 0.496150 0.250416 0.332841 0.936583 False Autogluon 0.008171 420500 [amt] 0.01 314572 0.005725 None
7 ExtraTreesGini None 0.992866 0.371296 0.354803 0.362862 0.860650 False Autogluon 0.008171 420500 [amt] 0.01 314572 0.005725 None
8 ExtraTreesEntr None 0.992911 0.373898 0.353137 0.363221 0.858746 False Autogluon 0.008171 420500 [amt] 0.01 314572 0.005725 None
9 NeuralNetFastAI None 0.993804 0.435315 0.276513 0.338200 0.929382 False Autogluon 0.008171 420500 [amt] 0.01 314572 0.005725 None
10 XGBoost None 0.994202 0.487514 0.249306 0.329904 0.939309 False Autogluon 0.008171 420500 [amt] 0.01 314572 0.005725 None
11 NeuralNetTorch None 0.994510 0.566071 0.176013 0.268530 0.952137 False Autogluon 0.008171 420500 [amt] 0.01 314572 0.005725 None
12 LightGBMLarge None 0.994275 0.000000 0.000000 0.000000 0.960304 False Autogluon 0.008171 420500 [amt] 0.01 314572 0.005725 None
13 WeightedEnsemble_L2 None 0.994510 0.566071 0.176013 0.268530 0.952137 False Autogluon 0.008171 420500 [amt] 0.01 314572 0.005725 None 
auto_amt_ver0503(df_train3, df_test, _df3_mean)
No path specified. Models will be saved in: "AutogluonModels/ag-20240520_053949/"
AutoGluon infers your prediction problem is: 'binary' (because only two unique label-values observed).
    If 'binary' is not the correct problem_type, please manually specify the problem_type parameter during predictor init (You may specify problem_type as one of: ['binary', 'multiclass', 'regression'])
model time acc pre rec f1 auc graph_based method throw_rate train_size train_cols train_frate test_size test_frate hyper_params
0 KNeighborsUnif None 0.988111 0.269002 0.626874 0.376459 0.881531 False Autogluon 0.015065 84100 [amt] 0.05 314572 0.005725 None
1 KNeighborsDist None 0.986839 0.239242 0.595780 0.341394 0.866643 False Autogluon 0.015065 84100 [amt] 0.05 314572 0.005725 None
2 LightGBMXT None 0.994275 0.000000 0.000000 0.000000 0.781026 False Autogluon 0.015065 84100 [amt] 0.05 314572 0.005725 None
3 LightGBM None 0.990857 0.327004 0.564131 0.414018 0.962248 False Autogluon 0.015065 84100 [amt] 0.05 314572 0.005725 None
4 RandomForestGini None 0.985933 0.221444 0.579123 0.320381 0.892199 False Autogluon 0.015065 84100 [amt] 0.05 314572 0.005725 None
5 RandomForestEntr None 0.985933 0.221444 0.579123 0.320381 0.892199 False Autogluon 0.015065 84100 [amt] 0.05 314572 0.005725 None
6 CatBoost None 0.989109 0.289889 0.622432 0.395554 0.947419 False Autogluon 0.015065 84100 [amt] 0.05 314572 0.005725 None
7 ExtraTreesGini None 0.986935 0.240853 0.595780 0.343031 0.901395 False Autogluon 0.015065 84100 [amt] 0.05 314572 0.005725 None
8 ExtraTreesEntr None 0.986881 0.239937 0.595780 0.342101 0.900649 False Autogluon 0.015065 84100 [amt] 0.05 314572 0.005725 None
9 NeuralNetFastAI None 0.992202 0.358630 0.459189 0.402727 0.892828 False Autogluon 0.015065 84100 [amt] 0.05 314572 0.005725 None
10 XGBoost None 0.989122 0.290080 0.621877 0.395620 0.947978 False Autogluon 0.015065 84100 [amt] 0.05 314572 0.005725 None
11 NeuralNetTorch None 0.989328 0.300205 0.649084 0.410536 0.956575 False Autogluon 0.015065 84100 [amt] 0.05 314572 0.005725 None
12 LightGBMLarge None 0.990857 0.327004 0.564131 0.414018 0.960620 False Autogluon 0.015065 84100 [amt] 0.05 314572 0.005725 None
13 WeightedEnsemble_L2 None 0.989675 0.307937 0.644087 0.416667 0.953609 False Autogluon 0.015065 84100 [amt] 0.05 314572 0.005725 None 
auto_amt_ver0503(df_train4, df_test, _df4_mean)
No path specified. Models will be saved in: "AutogluonModels/ag-20240520_054133/"
AutoGluon infers your prediction problem is: 'binary' (because only two unique label-values observed).
    If 'binary' is not the correct problem_type, please manually specify the problem_type parameter during predictor init (You may specify problem_type as one of: ['binary', 'multiclass', 'regression'])
model time acc pre rec f1 auc graph_based method throw_rate train_size train_cols train_frate test_size test_frate hyper_params
0 KNeighborsUnif None 0.982595 0.201106 0.686285 0.311061 0.900398 False Autogluon 0.016841 42050 [amt] 0.1 314572 0.005725 None
1 KNeighborsDist None 0.978968 0.164834 0.657413 0.263580 0.881212 False Autogluon 0.016841 42050 [amt] 0.1 314572 0.005725 None
2 LightGBMXT None 0.978984 0.161268 0.635758 0.257274 0.946402 False Autogluon 0.016841 42050 [amt] 0.1 314572 0.005725 None
3 LightGBM None 0.988766 0.286629 0.646308 0.397134 0.961948 False Autogluon 0.016841 42050 [amt] 0.1 314572 0.005725 None
4 RandomForestGini None 0.976641 0.146733 0.639645 0.238707 0.912443 False Autogluon 0.016841 42050 [amt] 0.1 314572 0.005725 None
5 RandomForestEntr None 0.976641 0.146733 0.639645 0.238707 0.912443 False Autogluon 0.016841 42050 [amt] 0.1 314572 0.005725 None
6 CatBoost None 0.984334 0.221946 0.692948 0.336207 0.954685 False Autogluon 0.016841 42050 [amt] 0.1 314572 0.005725 None
7 ExtraTreesGini None 0.978558 0.161833 0.656857 0.259686 0.922153 False Autogluon 0.016841 42050 [amt] 0.1 314572 0.005725 None
8 ExtraTreesEntr None 0.978565 0.162062 0.657968 0.260068 0.921562 False Autogluon 0.016841 42050 [amt] 0.1 314572 0.005725 None
9 NeuralNetFastAI None 0.978072 0.169241 0.724042 0.274353 0.933527 False Autogluon 0.016841 42050 [amt] 0.1 314572 0.005725 None
10 XGBoost None 0.986776 0.254322 0.677957 0.369888 0.948753 False Autogluon 0.016841 42050 [amt] 0.1 314572 0.005725 None
11 NeuralNetTorch None 0.984684 0.227363 0.698501 0.343060 0.958205 False Autogluon 0.016841 42050 [amt] 0.1 314572 0.005725 None
12 LightGBMLarge None 0.988766 0.286629 0.646308 0.397134 0.960951 False Autogluon 0.016841 42050 [amt] 0.1 314572 0.005725 None
13 WeightedEnsemble_L2 None 0.986868 0.257897 0.689062 0.375321 0.961334 False Autogluon 0.016841 42050 [amt] 0.1 314572 0.005725 None 
auto_amt_ver0503(df_train5, df_test, _df5_mean)
No path specified. Models will be saved in: "AutogluonModels/ag-20240520_054246/"
AutoGluon infers your prediction problem is: 'binary' (because only two unique label-values observed).
    If 'binary' is not the correct problem_type, please manually specify the problem_type parameter during predictor init (You may specify problem_type as one of: ['binary', 'multiclass', 'regression'])
model time acc pre rec f1 auc graph_based method throw_rate train_size train_cols train_frate test_size test_frate hyper_params
0 KNeighborsUnif None 0.963411 0.108540 0.747363 0.189551 0.923442 False Autogluon 0.017896 21025 [amt] 0.2 314572 0.005725 None
1 KNeighborsDist None 0.956846 0.090896 0.726263 0.161571 0.897693 False Autogluon 0.017896 21025 [amt] 0.2 314572 0.005725 None
2 LightGBMXT None 0.975347 0.154720 0.740700 0.255972 0.951622 False Autogluon 0.017896 21025 [amt] 0.2 314572 0.005725 None
3 LightGBM None 0.982532 0.205704 0.716824 0.319673 0.961749 False Autogluon 0.017896 21025 [amt] 0.2 314572 0.005725 None
4 RandomForestGini None 0.949843 0.078371 0.721266 0.141380 0.927522 False Autogluon 0.017896 21025 [amt] 0.2 314572 0.005725 None
5 RandomForestEntr None 0.949843 0.078371 0.721266 0.141380 0.927522 False Autogluon 0.017896 21025 [amt] 0.2 314572 0.005725 None
6 CatBoost None 0.981248 0.194635 0.725153 0.306897 0.962530 False Autogluon 0.017896 21025 [amt] 0.2 314572 0.005725 None
7 ExtraTreesGini None 0.953333 0.085378 0.736258 0.153012 0.933086 False Autogluon 0.017896 21025 [amt] 0.2 314572 0.005725 None
8 ExtraTreesEntr None 0.952920 0.084616 0.735702 0.151775 0.935070 False Autogluon 0.017896 21025 [amt] 0.2 314572 0.005725 None
9 NeuralNetFastAI None 0.973888 0.148217 0.750139 0.247527 0.948571 False Autogluon 0.017896 21025 [amt] 0.2 314572 0.005725 None
10 XGBoost None 0.981159 0.193917 0.725708 0.306053 0.954487 False Autogluon 0.017896 21025 [amt] 0.2 314572 0.005725 None
11 NeuralNetTorch None 0.975204 0.154952 0.747918 0.256718 0.954467 False Autogluon 0.017896 21025 [amt] 0.2 314572 0.005725 None
12 LightGBMLarge None 0.982532 0.205704 0.716824 0.319673 0.959817 False Autogluon 0.017896 21025 [amt] 0.2 314572 0.005725 None
13 WeightedEnsemble_L2 None 0.982821 0.209389 0.720711 0.324500 0.961966 False Autogluon 0.017896 21025 [amt] 0.2 314572 0.005725 None 
auto_amt_ver0503(df_train6, df_test, _df6_mean)
No path specified. Models will be saved in: "AutogluonModels/ag-20240520_054324/"
AutoGluon infers your prediction problem is: 'binary' (because only two unique label-values observed).
    If 'binary' is not the correct problem_type, please manually specify the problem_type parameter during predictor init (You may specify problem_type as one of: ['binary', 'multiclass', 'regression'])
model time acc pre rec f1 auc graph_based method throw_rate train_size train_cols train_frate test_size test_frate hyper_params
0 KNeighborsUnif None 0.940697 0.072537 0.794003 0.132931 0.933847 False Autogluon 0.018278 14017 [amt] 0.299993 314572 0.005725 None
1 KNeighborsDist None 0.932499 0.063441 0.784009 0.117383 0.904577 False Autogluon 0.018278 14017 [amt] 0.299993 314572 0.005725 None
2 LightGBMXT None 0.971243 0.136405 0.754581 0.231044 0.947721 False Autogluon 0.018278 14017 [amt] 0.299993 314572 0.005725 None
3 LightGBM None 0.964120 0.113510 0.773459 0.197968 0.961942 False Autogluon 0.018278 14017 [amt] 0.299993 314572 0.005725 None
4 RandomForestGini None 0.920498 0.053932 0.779012 0.100881 0.934945 False Autogluon 0.018278 14017 [amt] 0.299993 314572 0.005725 None
5 RandomForestEntr None 0.920498 0.053932 0.779012 0.100881 0.934945 False Autogluon 0.018278 14017 [amt] 0.299993 314572 0.005725 None
6 CatBoost None 0.969501 0.129434 0.755691 0.221013 0.959693 False Autogluon 0.018278 14017 [amt] 0.299993 314572 0.005725 None
7 ExtraTreesGini None 0.925063 0.057803 0.790117 0.107726 0.940287 False Autogluon 0.018278 14017 [amt] 0.299993 314572 0.005725 None
8 ExtraTreesEntr None 0.925763 0.058505 0.792893 0.108970 0.940250 False Autogluon 0.018278 14017 [amt] 0.299993 314572 0.005725 None
9 NeuralNetFastAI None 0.972080 0.139732 0.751805 0.235663 0.898952 False Autogluon 0.018278 14017 [amt] 0.299993 314572 0.005725 None
10 XGBoost None 0.961230 0.106459 0.780677 0.187367 0.962270 False Autogluon 0.018278 14017 [amt] 0.299993 314572 0.005725 None
11 NeuralNetTorch None 0.976031 0.159245 0.744586 0.262375 0.953301 False Autogluon 0.018278 14017 [amt] 0.299993 314572 0.005725 None
12 LightGBMLarge None 0.973211 0.144070 0.744586 0.241426 0.959773 False Autogluon 0.018278 14017 [amt] 0.299993 314572 0.005725 None
13 WeightedEnsemble_L2 None 0.973211 0.144070 0.744586 0.241426 0.959773 False Autogluon 0.018278 14017 [amt] 0.299993 314572 0.005725 None 
auto_amt_ver0503(df_train7, df_test, _df7_mean)
No path specified. Models will be saved in: "AutogluonModels/ag-20240520_054355/"
AutoGluon infers your prediction problem is: 'binary' (because only two unique label-values observed).
    If 'binary' is not the correct problem_type, please manually specify the problem_type parameter during predictor init (You may specify problem_type as one of: ['binary', 'multiclass', 'regression'])
model time acc pre rec f1 auc graph_based method throw_rate train_size train_cols train_frate test_size test_frate hyper_params
0 KNeighborsUnif None 0.914633 0.053725 0.837313 0.100971 0.939006 False Autogluon 0.018475 10512 [amt] 0.400019 314572 0.005725 None
1 KNeighborsDist None 0.905713 0.048228 0.825652 0.091132 0.911215 False Autogluon 0.018475 10512 [amt] 0.400019 314572 0.005725 None
2 LightGBMXT None 0.945599 0.080585 0.816768 0.146697 0.955586 False Autogluon 0.018475 10512 [amt] 0.400019 314572 0.005725 None
3 LightGBM None 0.934616 0.069347 0.838978 0.128105 0.961903 False Autogluon 0.018475 10512 [amt] 0.400019 314572 0.005725 None
4 RandomForestGini None 0.887406 0.040562 0.823987 0.077318 0.938095 False Autogluon 0.018475 10512 [amt] 0.400019 314572 0.005725 None
5 RandomForestEntr None 0.887406 0.040562 0.823987 0.077318 0.938095 False Autogluon 0.018475 10512 [amt] 0.400019 314572 0.005725 None
6 CatBoost None 0.946769 0.082009 0.813992 0.149006 0.958300 False Autogluon 0.018475 10512 [amt] 0.400019 314572 0.005725 None
7 ExtraTreesGini None 0.897276 0.045003 0.837868 0.085418 0.944199 False Autogluon 0.018475 10512 [amt] 0.400019 314572 0.005725 None
8 ExtraTreesEntr None 0.896577 0.044737 0.838423 0.084941 0.943396 False Autogluon 0.018475 10512 [amt] 0.400019 314572 0.005725 None
9 NeuralNetFastAI None 0.970783 0.134520 0.755136 0.228360 0.901838 False Autogluon 0.018475 10512 [amt] 0.400019 314572 0.005725 None
10 XGBoost None 0.932667 0.067681 0.842310 0.125294 0.961863 False Autogluon 0.018475 10512 [amt] 0.400019 314572 0.005725 None
11 NeuralNetTorch None 0.938733 0.072435 0.821766 0.133135 0.955618 False Autogluon 0.018475 10512 [amt] 0.400019 314572 0.005725 None
12 LightGBMLarge None 0.932311 0.067767 0.848418 0.125508 0.960788 False Autogluon 0.018475 10512 [amt] 0.400019 314572 0.005725 None
13 WeightedEnsemble_L2 None 0.934616 0.069347 0.838978 0.128105 0.961903 False Autogluon 0.018475 10512 [amt] 0.400019 314572 0.005725 None 
auto_amt_ver0503(df_train8, df_test, _df8_mean)
No path specified. Models will be saved in: "AutogluonModels/ag-20240520_054431/"
AutoGluon infers your prediction problem is: 'binary' (because only two unique label-values observed).
    If 'binary' is not the correct problem_type, please manually specify the problem_type parameter during predictor init (You may specify problem_type as one of: ['binary', 'multiclass', 'regression'])
model time acc pre rec f1 auc graph_based method throw_rate train_size train_cols train_frate test_size test_frate hyper_params
0 KNeighborsUnif None 0.879996 0.040024 0.868406 0.076520 0.938083 False Autogluon 0.018595 8410 [amt] 0.5 314572 0.005725 None
1 KNeighborsDist None 0.876337 0.038557 0.860633 0.073808 0.907000 False Autogluon 0.018595 8410 [amt] 0.5 314572 0.005725 None
2 LightGBMXT None 0.883591 0.041820 0.882288 0.079855 0.952322 False Autogluon 0.018595 8410 [amt] 0.5 314572 0.005725 None
3 LightGBM None 0.908285 0.052093 0.873404 0.098322 0.960795 False Autogluon 0.018595 8410 [amt] 0.5 314572 0.005725 None
4 RandomForestGini None 0.860633 0.034338 0.860633 0.066040 0.939192 False Autogluon 0.018595 8410 [amt] 0.5 314572 0.005725 None
5 RandomForestEntr None 0.860633 0.034338 0.860633 0.066040 0.939192 False Autogluon 0.018595 8410 [amt] 0.5 314572 0.005725 None
6 CatBoost None 0.900090 0.048243 0.878401 0.091464 0.961486 False Autogluon 0.018595 8410 [amt] 0.5 314572 0.005725 None
7 ExtraTreesGini None 0.866968 0.036225 0.868406 0.069548 0.942842 False Autogluon 0.018595 8410 [amt] 0.5 314572 0.005725 None
8 ExtraTreesEntr None 0.868199 0.036466 0.866185 0.069985 0.942820 False Autogluon 0.018595 8410 [amt] 0.5 314572 0.005725 None
9 NeuralNetFastAI None 0.966326 0.118535 0.758468 0.205028 0.898449 False Autogluon 0.018595 8410 [amt] 0.5 314572 0.005725 None
10 XGBoost None 0.903482 0.049752 0.876180 0.094158 0.961121 False Autogluon 0.018595 8410 [amt] 0.5 314572 0.005725 None
11 NeuralNetTorch None 0.842033 0.032196 0.915047 0.062203 0.950496 False Autogluon 0.018595 8410 [amt] 0.5 314572 0.005725 None
12 LightGBMLarge None 0.892727 0.045191 0.881177 0.085972 0.960270 False Autogluon 0.018595 8410 [amt] 0.5 314572 0.005725 None
13 WeightedEnsemble_L2 None 0.849138 0.033570 0.912271 0.064758 0.950978 False Autogluon 0.018595 8410 [amt] 0.5 314572 0.005725 None

amt 80 미만 잘 잡는지 확인용..

df_80 = df_test[df_test['amt'] <= 80]
df_80.shape, df_80.is_fraud.mean()
((231011, 22), 0.0016665873053664112)
_df1_ = pd.concat([df_train1, df_80])
_df2_ = pd.concat([df_train2, df_80])
_df3_ = pd.concat([df_train3, df_80])
_df4_ = pd.concat([df_train4, df_80])
_df5_ = pd.concat([df_train5, df_80])
_df6_ = pd.concat([df_train6, df_80])
_df7_ = pd.concat([df_train7, df_80])
_df8_ = pd.concat([df_train8, df_80])
_df1_mean_ = _df1_.is_fraud.mean()
_df2_mean_ = _df2_.is_fraud.mean()
_df3_mean_ = _df3_.is_fraud.mean()
_df4_mean_ = _df4_.is_fraud.mean()
_df5_mean_ = _df5_.is_fraud.mean()
_df6_mean_ = _df6_.is_fraud.mean()
_df7_mean_ = _df7_.is_fraud.mean()
_df8_mean_ = _df8_.is_fraud.mean()
auto_amt_ver0503(df_train1, df_80, _df1_mean_)
No path specified. Models will be saved in: "AutogluonModels/ag-20240520_094408/"
AutoGluon infers your prediction problem is: 'binary' (because only two unique label-values observed).
    If 'binary' is not the correct problem_type, please manually specify the problem_type parameter during predictor init (You may specify problem_type as one of: ['binary', 'multiclass', 'regression'])
auto_amt_ver0503(df_train2, df_80, _df2_mean_)
No path specified. Models will be saved in: "AutogluonModels/ag-20240520_064424/"
AutoGluon infers your prediction problem is: 'binary' (because only two unique label-values observed).
    If 'binary' is not the correct problem_type, please manually specify the problem_type parameter during predictor init (You may specify problem_type as one of: ['binary', 'multiclass', 'regression'])
model time acc pre rec f1 auc graph_based method throw_rate train_size train_cols train_frate test_size test_frate hyper_params
0 KNeighborsUnif None 0.998333 0.0 0.0 0.0 0.537194 False Autogluon 0.007045 420500 [amt] 0.01 231011 0.001667 None
1 KNeighborsDist None 0.998333 0.0 0.0 0.0 0.537194 False Autogluon 0.007045 420500 [amt] 0.01 231011 0.001667 None
2 LightGBMXT None 0.998333 0.0 0.0 0.0 0.619651 False Autogluon 0.007045 420500 [amt] 0.01 231011 0.001667 None
3 LightGBM None 0.998333 0.0 0.0 0.0 0.885680 False Autogluon 0.007045 420500 [amt] 0.01 231011 0.001667 None
4 RandomForestGini None 0.998333 0.0 0.0 0.0 0.631076 False Autogluon 0.007045 420500 [amt] 0.01 231011 0.001667 None
5 RandomForestEntr None 0.998333 0.0 0.0 0.0 0.631076 False Autogluon 0.007045 420500 [amt] 0.01 231011 0.001667 None
6 CatBoost None 0.998333 0.0 0.0 0.0 0.854812 False Autogluon 0.007045 420500 [amt] 0.01 231011 0.001667 None
7 ExtraTreesGini None 0.998333 0.0 0.0 0.0 0.631076 False Autogluon 0.007045 420500 [amt] 0.01 231011 0.001667 None
8 ExtraTreesEntr None 0.998333 0.0 0.0 0.0 0.631076 False Autogluon 0.007045 420500 [amt] 0.01 231011 0.001667 None
9 NeuralNetFastAI None 0.998333 0.0 0.0 0.0 0.746475 False Autogluon 0.007045 420500 [amt] 0.01 231011 0.001667 None
10 XGBoost None 0.998333 0.0 0.0 0.0 0.826658 False Autogluon 0.007045 420500 [amt] 0.01 231011 0.001667 None
11 NeuralNetTorch None 0.998333 0.0 0.0 0.0 0.858653 False Autogluon 0.007045 420500 [amt] 0.01 231011 0.001667 None
12 LightGBMLarge None 0.998333 0.0 0.0 0.0 0.885310 False Autogluon 0.007045 420500 [amt] 0.01 231011 0.001667 None
13 WeightedEnsemble_L2 None 0.998333 0.0 0.0 0.0 0.858653 False Autogluon 0.007045 420500 [amt] 0.01 231011 0.001667 None 
auto_amt_ver0503(df_train3, df_80, _df3_mean_)
No path specified. Models will be saved in: "AutogluonModels/ag-20240520_064813/"
AutoGluon infers your prediction problem is: 'binary' (because only two unique label-values observed).
    If 'binary' is not the correct problem_type, please manually specify the problem_type parameter during predictor init (You may specify problem_type as one of: ['binary', 'multiclass', 'regression'])
model time acc pre rec f1 auc graph_based method throw_rate train_size train_cols train_frate test_size test_frate hyper_params
0 KNeighborsUnif None 0.997074 0.010101 0.007792 0.008798 0.610442 False Autogluon 0.014566 84100 [amt] 0.05 231011 0.001667 None
1 KNeighborsDist None 0.996883 0.011662 0.010390 0.010989 0.592382 False Autogluon 0.014566 84100 [amt] 0.05 231011 0.001667 None
2 LightGBMXT None 0.998333 0.000000 0.000000 0.000000 0.667233 False Autogluon 0.014566 84100 [amt] 0.05 231011 0.001667 None
3 LightGBM None 0.998333 0.000000 0.000000 0.000000 0.890164 False Autogluon 0.014566 84100 [amt] 0.05 231011 0.001667 None
4 RandomForestGini None 0.996953 0.009231 0.007792 0.008451 0.665573 False Autogluon 0.014566 84100 [amt] 0.05 231011 0.001667 None
5 RandomForestEntr None 0.996953 0.009231 0.007792 0.008451 0.665573 False Autogluon 0.014566 84100 [amt] 0.05 231011 0.001667 None
6 CatBoost None 0.998333 0.000000 0.000000 0.000000 0.849362 False Autogluon 0.014566 84100 [amt] 0.05 231011 0.001667 None
7 ExtraTreesGini None 0.997186 0.011070 0.007792 0.009146 0.679732 False Autogluon 0.014566 84100 [amt] 0.05 231011 0.001667 None
8 ExtraTreesEntr None 0.997186 0.011070 0.007792 0.009146 0.677500 False Autogluon 0.014566 84100 [amt] 0.05 231011 0.001667 None
9 NeuralNetFastAI None 0.998333 0.000000 0.000000 0.000000 0.593622 False Autogluon 0.014566 84100 [amt] 0.05 231011 0.001667 None
10 XGBoost None 0.998333 0.000000 0.000000 0.000000 0.854089 False Autogluon 0.014566 84100 [amt] 0.05 231011 0.001667 None
11 NeuralNetTorch None 0.998333 0.000000 0.000000 0.000000 0.862387 False Autogluon 0.014566 84100 [amt] 0.05 231011 0.001667 None
12 LightGBMLarge None 0.998333 0.000000 0.000000 0.000000 0.885963 False Autogluon 0.014566 84100 [amt] 0.05 231011 0.001667 None
13 WeightedEnsemble_L2 None 0.998333 0.000000 0.000000 0.000000 0.869034 False Autogluon 0.014566 84100 [amt] 0.05 231011 0.001667 None 
auto_amt_ver0503(df_train4, df_80, _df4_mean_)
No path specified. Models will be saved in: "AutogluonModels/ag-20240520_065005/"
AutoGluon infers your prediction problem is: 'binary' (because only two unique label-values observed).
    If 'binary' is not the correct problem_type, please manually specify the problem_type parameter during predictor init (You may specify problem_type as one of: ['binary', 'multiclass', 'regression'])
model time acc pre rec f1 auc graph_based method throw_rate train_size train_cols train_frate test_size test_frate hyper_params
0 KNeighborsUnif None 0.993780 0.006567 0.018182 0.009649 0.663976 False Autogluon 0.016809 42050 [amt] 0.1 231011 0.001667 None
1 KNeighborsDist None 0.991113 0.008255 0.036364 0.013455 0.618455 False Autogluon 0.016809 42050 [amt] 0.1 231011 0.001667 None
2 LightGBMXT None 0.998333 0.000000 0.000000 0.000000 0.865731 False Autogluon 0.016809 42050 [amt] 0.1 231011 0.001667 None
3 LightGBM None 0.998333 0.000000 0.000000 0.000000 0.888579 False Autogluon 0.016809 42050 [amt] 0.1 231011 0.001667 None
4 RandomForestGini None 0.989940 0.009114 0.046753 0.015254 0.728206 False Autogluon 0.016809 42050 [amt] 0.1 231011 0.001667 None
5 RandomForestEntr None 0.989940 0.009114 0.046753 0.015254 0.728206 False Autogluon 0.016809 42050 [amt] 0.1 231011 0.001667 None
6 CatBoost None 0.998333 0.000000 0.000000 0.000000 0.877423 False Autogluon 0.016809 42050 [amt] 0.1 231011 0.001667 None
7 ExtraTreesGini None 0.990723 0.008939 0.041558 0.014713 0.758804 False Autogluon 0.016809 42050 [amt] 0.1 231011 0.001667 None
8 ExtraTreesEntr None 0.990710 0.008375 0.038961 0.013787 0.754776 False Autogluon 0.016809 42050 [amt] 0.1 231011 0.001667 None
9 NeuralNetFastAI None 0.998333 0.000000 0.000000 0.000000 0.778624 False Autogluon 0.016809 42050 [amt] 0.1 231011 0.001667 None
10 XGBoost None 0.998333 0.000000 0.000000 0.000000 0.857305 False Autogluon 0.016809 42050 [amt] 0.1 231011 0.001667 None
11 NeuralNetTorch None 0.998333 0.000000 0.000000 0.000000 0.875832 False Autogluon 0.016809 42050 [amt] 0.1 231011 0.001667 None
12 LightGBMLarge None 0.998333 0.000000 0.000000 0.000000 0.883690 False Autogluon 0.016809 42050 [amt] 0.1 231011 0.001667 None
13 WeightedEnsemble_L2 None 0.998333 0.000000 0.000000 0.000000 0.886679 False Autogluon 0.016809 42050 [amt] 0.1 231011 0.001667 None 
auto_amt_ver0503(df_train5, df_80, _df5_mean_)
No path specified. Models will be saved in: "AutogluonModels/ag-20240520_065116/"
AutoGluon infers your prediction problem is: 'binary' (because only two unique label-values observed).
    If 'binary' is not the correct problem_type, please manually specify the problem_type parameter during predictor init (You may specify problem_type as one of: ['binary', 'multiclass', 'regression'])
model time acc pre rec f1 auc graph_based method throw_rate train_size train_cols train_frate test_size test_frate hyper_params
0 KNeighborsUnif None 0.974651 0.009679 0.140260 0.018109 0.762008 False Autogluon 0.018212 21025 [amt] 0.2 231011 0.001667 None
1 KNeighborsDist None 0.968343 0.007955 0.145455 0.015084 0.686752 False Autogluon 0.018212 21025 [amt] 0.2 231011 0.001667 None
2 LightGBMXT None 0.998333 0.000000 0.000000 0.000000 0.868728 False Autogluon 0.018212 21025 [amt] 0.2 231011 0.001667 None
3 LightGBM None 0.998333 0.000000 0.000000 0.000000 0.888796 False Autogluon 0.018212 21025 [amt] 0.2 231011 0.001667 None
4 RandomForestGini None 0.961876 0.008405 0.187013 0.016088 0.790183 False Autogluon 0.018212 21025 [amt] 0.2 231011 0.001667 None
5 RandomForestEntr None 0.961876 0.008405 0.187013 0.016088 0.790183 False Autogluon 0.018212 21025 [amt] 0.2 231011 0.001667 None
6 CatBoost None 0.998333 0.000000 0.000000 0.000000 0.891512 False Autogluon 0.018212 21025 [amt] 0.2 231011 0.001667 None
7 ExtraTreesGini None 0.964097 0.009063 0.189610 0.017299 0.804425 False Autogluon 0.018212 21025 [amt] 0.2 231011 0.001667 None
8 ExtraTreesEntr None 0.963664 0.008831 0.187013 0.016866 0.809713 False Autogluon 0.018212 21025 [amt] 0.2 231011 0.001667 None
9 NeuralNetFastAI None 0.998333 0.000000 0.000000 0.000000 0.855702 False Autogluon 0.018212 21025 [amt] 0.2 231011 0.001667 None
10 XGBoost None 0.998333 0.000000 0.000000 0.000000 0.863929 False Autogluon 0.018212 21025 [amt] 0.2 231011 0.001667 None
11 NeuralNetTorch None 0.998333 0.000000 0.000000 0.000000 0.862611 False Autogluon 0.018212 21025 [amt] 0.2 231011 0.001667 None
12 LightGBMLarge None 0.998333 0.000000 0.000000 0.000000 0.884554 False Autogluon 0.018212 21025 [amt] 0.2 231011 0.001667 None
13 WeightedEnsemble_L2 None 0.998333 0.000000 0.000000 0.000000 0.889678 False Autogluon 0.018212 21025 [amt] 0.2 231011 0.001667 None 
auto_amt_ver0503(df_train6, df_80, _df6_mean_)
No path specified. Models will be saved in: "AutogluonModels/ag-20240520_065157/"
AutoGluon infers your prediction problem is: 'binary' (because only two unique label-values observed).
    If 'binary' is not the correct problem_type, please manually specify the problem_type parameter during predictor init (You may specify problem_type as one of: ['binary', 'multiclass', 'regression'])
model time acc pre rec f1 auc graph_based method throw_rate train_size train_cols train_frate test_size test_frate hyper_params
0 KNeighborsUnif None 0.949574 0.009408 0.280519 0.018205 0.808472 False Autogluon 0.018733 14017 [amt] 0.299993 231011 0.001667 None
1 KNeighborsDist None 0.942470 0.008381 0.285714 0.016284 0.725144 False Autogluon 0.018733 14017 [amt] 0.299993 231011 0.001667 None
2 LightGBMXT None 0.998333 0.000000 0.000000 0.000000 0.860889 False Autogluon 0.018733 14017 [amt] 0.299993 231011 0.001667 None
3 LightGBM None 0.985395 0.014300 0.114286 0.025419 0.890098 False Autogluon 0.018733 14017 [amt] 0.299993 231011 0.001667 None
4 RandomForestGini None 0.930609 0.007616 0.314286 0.014872 0.818151 False Autogluon 0.018733 14017 [amt] 0.299993 231011 0.001667 None
5 RandomForestEntr None 0.930609 0.007616 0.314286 0.014872 0.818151 False Autogluon 0.018733 14017 [amt] 0.299993 231011 0.001667 None
6 CatBoost None 0.993758 0.010195 0.028571 0.015027 0.883878 False Autogluon 0.018733 14017 [amt] 0.299993 231011 0.001667 None
7 ExtraTreesGini None 0.932977 0.008337 0.332468 0.016265 0.830028 False Autogluon 0.018733 14017 [amt] 0.299993 231011 0.001667 None
8 ExtraTreesEntr None 0.934016 0.008404 0.329870 0.016390 0.831543 False Autogluon 0.018733 14017 [amt] 0.299993 231011 0.001667 None
9 NeuralNetFastAI None 0.998333 0.000000 0.000000 0.000000 0.604243 False Autogluon 0.018733 14017 [amt] 0.299993 231011 0.001667 None
10 XGBoost None 0.982451 0.014811 0.145455 0.026884 0.889313 False Autogluon 0.018733 14017 [amt] 0.299993 231011 0.001667 None
11 NeuralNetTorch None 0.998333 0.000000 0.000000 0.000000 0.861303 False Autogluon 0.018733 14017 [amt] 0.299993 231011 0.001667 None
12 LightGBMLarge None 0.989671 0.013612 0.072727 0.022932 0.885505 False Autogluon 0.018733 14017 [amt] 0.299993 231011 0.001667 None
13 WeightedEnsemble_L2 None 0.989671 0.013612 0.072727 0.022932 0.885505 False Autogluon 0.018733 14017 [amt] 0.299993 231011 0.001667 None 
auto_amt_ver0503(df_train7, df_80, _df7_mean_)
No path specified. Models will be saved in: "AutogluonModels/ag-20240520_065229/"
AutoGluon infers your prediction problem is: 'binary' (because only two unique label-values observed).
    If 'binary' is not the correct problem_type, please manually specify the problem_type parameter during predictor init (You may specify problem_type as one of: ['binary', 'multiclass', 'regression'])
model time acc pre rec f1 auc graph_based method throw_rate train_size train_cols train_frate test_size test_frate hyper_params
0 KNeighborsUnif None 0.916411 0.008416 0.420779 0.016502 0.828619 False Autogluon 0.019004 10512 [amt] 0.400019 231011 0.001667 None
1 KNeighborsDist None 0.912039 0.007753 0.407792 0.015218 0.762284 False Autogluon 0.019004 10512 [amt] 0.400019 231011 0.001667 None
2 LightGBMXT None 0.963513 0.013546 0.290909 0.025887 0.876453 False Autogluon 0.019004 10512 [amt] 0.400019 231011 0.001667 None
3 LightGBM None 0.948557 0.012878 0.394805 0.024943 0.889180 False Autogluon 0.019004 10512 [amt] 0.400019 231011 0.001667 None
4 RandomForestGini None 0.893364 0.006753 0.431169 0.013298 0.833487 False Autogluon 0.019004 10512 [amt] 0.400019 231011 0.001667 None
5 RandomForestEntr None 0.893364 0.006753 0.431169 0.013298 0.833487 False Autogluon 0.019004 10512 [amt] 0.400019 231011 0.001667 None
6 CatBoost None 0.965032 0.013532 0.277922 0.025808 0.880542 False Autogluon 0.019004 10512 [amt] 0.400019 231011 0.001667 None
7 ExtraTreesGini None 0.901057 0.007668 0.454545 0.015082 0.846933 False Autogluon 0.019004 10512 [amt] 0.400019 231011 0.001667 None
8 ExtraTreesEntr None 0.900381 0.007702 0.459740 0.015150 0.843639 False Autogluon 0.019004 10512 [amt] 0.400019 231011 0.001667 None
9 NeuralNetFastAI None 0.998333 0.000000 0.000000 0.000000 0.604243 False Autogluon 0.019004 10512 [amt] 0.400019 231011 0.001667 None
10 XGBoost None 0.945548 0.012786 0.415584 0.024808 0.886951 False Autogluon 0.019004 10512 [amt] 0.400019 231011 0.001667 None
11 NeuralNetTorch None 0.950418 0.011561 0.340260 0.022363 0.868049 False Autogluon 0.019004 10512 [amt] 0.400019 231011 0.001667 None
12 LightGBMLarge None 0.945418 0.013453 0.438961 0.026106 0.887662 False Autogluon 0.019004 10512 [amt] 0.400019 231011 0.001667 None
13 WeightedEnsemble_L2 None 0.948557 0.012878 0.394805 0.024943 0.889180 False Autogluon 0.019004 10512 [amt] 0.400019 231011 0.001667 None 
auto_amt_ver0503(df_train8, df_80, _df8_mean_)
No path specified. Models will be saved in: "AutogluonModels/ag-20240520_065302/"
AutoGluon infers your prediction problem is: 'binary' (because only two unique label-values observed).
    If 'binary' is not the correct problem_type, please manually specify the problem_type parameter during predictor init (You may specify problem_type as one of: ['binary', 'multiclass', 'regression'])
model time acc pre rec f1 auc graph_based method throw_rate train_size train_cols train_frate test_size test_frate hyper_params
0 KNeighborsUnif None 0.876305 0.007306 0.542857 0.014417 0.835443 False Autogluon 0.019171 8410 [amt] 0.5 231011 0.001667 None
1 KNeighborsDist None 0.879227 0.007165 0.519481 0.014134 0.761894 False Autogluon 0.019171 8410 [amt] 0.5 231011 0.001667 None
2 LightGBMXT None 0.878248 0.008225 0.602597 0.016229 0.861540 False Autogluon 0.019171 8410 [amt] 0.5 231011 0.001667 None
3 LightGBM None 0.916974 0.010827 0.540260 0.021229 0.887029 False Autogluon 0.019171 8410 [amt] 0.5 231011 0.001667 None
4 RandomForestGini None 0.864556 0.006673 0.542857 0.013183 0.837880 False Autogluon 0.019171 8410 [amt] 0.5 231011 0.001667 None
5 RandomForestEntr None 0.864556 0.006673 0.542857 0.013183 0.837880 False Autogluon 0.019171 8410 [amt] 0.5 231011 0.001667 None
6 CatBoost None 0.900442 0.009840 0.589610 0.019358 0.888977 False Autogluon 0.019171 8410 [amt] 0.5 231011 0.001667 None
7 ExtraTreesGini None 0.868123 0.007015 0.555844 0.013854 0.841006 False Autogluon 0.019171 8410 [amt] 0.5 231011 0.001667 None
8 ExtraTreesEntr None 0.869738 0.007003 0.548052 0.013830 0.842724 False Autogluon 0.019171 8410 [amt] 0.5 231011 0.001667 None
9 NeuralNetFastAI None 0.998333 0.000000 0.000000 0.000000 0.604243 False Autogluon 0.019171 8410 [amt] 0.5 231011 0.001667 None
10 XGBoost None 0.907489 0.010131 0.563636 0.019904 0.887128 False Autogluon 0.019171 8410 [amt] 0.5 231011 0.001667 None
11 NeuralNetTorch None 0.842518 0.007579 0.719481 0.015000 0.859309 False Autogluon 0.019171 8410 [amt] 0.5 231011 0.001667 None
12 LightGBMLarge None 0.890689 0.009081 0.597403 0.017890 0.882991 False Autogluon 0.019171 8410 [amt] 0.5 231011 0.001667 None
13 WeightedEnsemble_L2 None 0.848488 0.007736 0.706494 0.015305 0.861127 False Autogluon 0.019171 8410 [amt] 0.5 231011 0.001667 None

다른 것 추가

def auto_amt_ver0520(df_tr, df_tst, _df_mean):
    df_tr = df_tr[['trans_date_trans_time',  'merchant', 'category', 'amt',
       'first', 'last', 'gender', 'street', 'city', 'state', 'zip', 'lat',
       'long', 'city_pop', 'job', 'dob', 'trans_num', 'unix_time', 'merch_lat',
       'merch_long', 'is_fraud']]
    df_tst = df_tst[['trans_date_trans_time',  'merchant', 'category', 'amt',
       'first', 'last', 'gender', 'street', 'city', 'state', 'zip', 'lat',
       'long', 'city_pop', 'job', 'dob', 'trans_num', 'unix_time', 'merch_lat',
       'merch_long', 'is_fraud']]     
    tr = TabularDataset(df_tr)
    tst = TabularDataset(df_tst)
    predictr = TabularPredictor(label="is_fraud", verbosity=1)
    t1 = time.time()
    predictr.fit(tr)
    t2 = time.time()
    time_diff = t2 - t1
    models = predictr._trainer.model_graph.nodes
    results = []
    for model_name in models:
    # 모델 평가
        eval_result = predictr.evaluate(tst, model=model_name)

    # 결과를 데이터프레임에 추가
        results.append({'model': model_name, 
                        'acc': eval_result['accuracy'], 
                        'pre': eval_result['precision'], 
                        'rec': eval_result['recall'], 
                        'f1': eval_result['f1'], 
                        'auc': eval_result['roc_auc']})
        
    model = []
    time_diff = []
    acc = []
    pre = []
    rec = []
    f1 = [] 
    auc = [] 
    graph_based = []
    method = [] 
    throw_rate = [] 
    train_size = []
    train_cols = []
    train_frate = []
    test_size = []
    test_frate = []
    hyper_params = [] 
    
    for result in results:
        model_name = result['model']
        model.append(model_name)
        time_diff.append(None)  # 각 모델별로 학습한 시간을 나타내고 싶은데 잘 안됨
        acc.append(result['acc']) 
        pre.append(result['pre'])
        rec.append(result['rec'])
        f1.append(result['f1'])
        auc.append(result['auc'])
        graph_based.append(False) 
        method.append('Autogluon') 
        throw_rate.append(_df_mean)
        train_size.append(len(tr))
        train_cols.append([col for col in tr.columns if col != 'is_fraud'])
        train_frate.append(tr.is_fraud.mean())
        test_size.append(len(tst))
        test_frate.append(tst.is_fraud.mean())
        hyper_params.append(None)
        
    df_results = pd.DataFrame(dict(
        model=model,
        time=time_diff,
        acc=acc,
        pre=pre,
        rec=rec,
        f1=f1,
        auc=auc,
        graph_based=graph_based,
        method=method,
        throw_rate=throw_rate,  
        train_size=train_size,
        train_cols=train_cols,
        train_frate=train_frate,
        test_size=test_size,
        test_frate=test_frate,
        hyper_params=hyper_params
    ))    
    ymdhms = datetime.datetime.fromtimestamp(time.time()).strftime('%Y%m%d-%H%M%S') 
    df_results.to_csv(f'../results2/{ymdhms}-Autogluon.csv',index=False)
    return df_results
auto_amt_ver0520(df_train1, df_test, _df1_mean)
No path specified. Models will be saved in: "AutogluonModels/ag-20240520_094915/"
AutoGluon infers your prediction problem is: 'binary' (because only two unique label-values observed).
    If 'binary' is not the correct problem_type, please manually specify the problem_type parameter during predictor init (You may specify problem_type as one of: ['binary', 'multiclass', 'regression'])
model time acc pre rec f1 auc graph_based method throw_rate train_size train_cols train_frate test_size test_frate hyper_params
0 KNeighborsUnif None 0.997854 0.774099 0.882843 0.824903 0.996391 False Autogluon 0.005728 734003 [trans_date_trans_time, merchant, category, amt, first, last, gender, street, city, state, zip, lat, long, city_pop, job, dob, trans_num, unix_time, merch_lat, merch_long] 0.005729 314572 0.005725 None
1 KNeighborsDist None 0.998760 0.850472 0.950583 0.897745 0.996693 False Autogluon 0.005728 734003 [trans_date_trans_time, merchant, category, amt, first, last, gender, street, city, state, zip, lat, long, city_pop, job, dob, trans_num, unix_time, merch_lat, merch_long] 0.005729 314572 0.005725 None
2 LightGBMXT None 0.997641 0.940150 0.627984 0.752996 0.995792 False Autogluon 0.005728 734003 [trans_date_trans_time, merchant, category, amt, first, last, gender, street, city, state, zip, lat, long, city_pop, job, dob, trans_num, unix_time, merch_lat, merch_long] 0.005729 314572 0.005725 None
3 LightGBM None 0.997266 0.838246 0.647418 0.730576 0.992670 False Autogluon 0.005728 734003 [trans_date_trans_time, merchant, category, amt, first, last, gender, street, city, state, zip, lat, long, city_pop, job, dob, trans_num, unix_time, merch_lat, merch_long] 0.005729 314572 0.005725 None
4 RandomForestGini None 0.997965 0.955294 0.676291 0.791938 0.998141 False Autogluon 0.005728 734003 [trans_date_trans_time, merchant, category, amt, first, last, gender, street, city, state, zip, lat, long, city_pop, job, dob, trans_num, unix_time, merch_lat, merch_long] 0.005729 314572 0.005725 None
5 RandomForestEntr None 0.998163 0.978857 0.694059 0.812216 0.998040 False Autogluon 0.005728 734003 [trans_date_trans_time, merchant, category, amt, first, last, gender, street, city, state, zip, lat, long, city_pop, job, dob, trans_num, unix_time, merch_lat, merch_long] 0.005729 314572 0.005725 None
6 CatBoost None 0.998404 0.940937 0.769572 0.846671 0.983846 False Autogluon 0.005728 734003 [trans_date_trans_time, merchant, category, amt, first, last, gender, street, city, state, zip, lat, long, city_pop, job, dob, trans_num, unix_time, merch_lat, merch_long] 0.005729 314572 0.005725 None
7 ExtraTreesGini None 0.997603 0.995270 0.584120 0.736179 0.999165 False Autogluon 0.005728 734003 [trans_date_trans_time, merchant, category, amt, first, last, gender, street, city, state, zip, lat, long, city_pop, job, dob, trans_num, unix_time, merch_lat, merch_long] 0.005729 314572 0.005725 None
8 ExtraTreesEntr None 0.997597 0.994324 0.583565 0.735479 0.999239 False Autogluon 0.005728 734003 [trans_date_trans_time, merchant, category, amt, first, last, gender, street, city, state, zip, lat, long, city_pop, job, dob, trans_num, unix_time, merch_lat, merch_long] 0.005729 314572 0.005725 None
9 NeuralNetFastAI None 0.998992 0.944311 0.875625 0.908672 0.997173 False Autogluon 0.005728 734003 [trans_date_trans_time, merchant, category, amt, first, last, gender, street, city, state, zip, lat, long, city_pop, job, dob, trans_num, unix_time, merch_lat, merch_long] 0.005729 314572 0.005725 None
10 XGBoost None 0.997253 0.840727 0.641866 0.727960 0.980501 False Autogluon 0.005728 734003 [trans_date_trans_time, merchant, category, amt, first, last, gender, street, city, state, zip, lat, long, city_pop, job, dob, trans_num, unix_time, merch_lat, merch_long] 0.005729 314572 0.005725 None
11 NeuralNetTorch None 0.997619 0.865786 0.691283 0.768756 0.993116 False Autogluon 0.005728 734003 [trans_date_trans_time, merchant, category, amt, first, last, gender, street, city, state, zip, lat, long, city_pop, job, dob, trans_num, unix_time, merch_lat, merch_long] 0.005729 314572 0.005725 None
12 LightGBMLarge None 0.997543 0.858939 0.682954 0.760903 0.990375 False Autogluon 0.005728 734003 [trans_date_trans_time, merchant, category, amt, first, last, gender, street, city, state, zip, lat, long, city_pop, job, dob, trans_num, unix_time, merch_lat, merch_long] 0.005729 314572 0.005725 None
13 WeightedEnsemble_L2 None 0.999399 0.975797 0.917823 0.945923 0.999333 False Autogluon 0.005728 734003 [trans_date_trans_time, merchant, category, amt, first, last, gender, street, city, state, zip, lat, long, city_pop, job, dob, trans_num, unix_time, merch_lat, merch_long] 0.005729 314572 0.005725 None 
Xcolumns = ['category', 'amt','gender', 'street', 'city', 'state', 'zip', 'lat', 'long', 'city_pop', 'job', 'unix_time', 'is_fraud']
df_tr = df_train1[Xcolumns]
df_tst = df_test[Xcolumns]
tr = TabularDataset(df_tr)
tst = TabularDataset(df_tst)
predictr = TabularPredictor(label="is_fraud", verbosity=1)
No path specified. Models will be saved in: "AutogluonModels/ag-20240520_103239/"
predictr.fit(tr)
AutoGluon infers your prediction problem is: 'binary' (because only two unique label-values observed).
    If 'binary' is not the correct problem_type, please manually specify the problem_type parameter during predictor init (You may specify problem_type as one of: ['binary', 'multiclass', 'regression'])
<autogluon.tabular.predictor.predictor.TabularPredictor at 0x7f6099bb3400>
yyhat_prob = predictr.predict_proba(tst).iloc[:,-1]
df_tst_compact=tst.assign(yyhat_prob = yyhat_prob).loc[:,['amt','is_fraud','yyhat_prob']]
df_tst_compact[df_tst_compact.amt<80]
amt is_fraud yyhat_prob
0 7.53 0 0.001258
1 3.79 0 0.001710
2 59.07 0 0.002354
3 25.58 0 0.000962
5 20.59 0 0.004030
... ... ... ...
314538 19.68 1 0.005087
314558 12.43 1 0.064472
314563 20.51 1 0.003546
314566 17.83 1 0.003011
314571 12.57 1 0.995636

230999 rows × 3 columns

df_tst_compact[
    (df_tst_compact.amt<80) & (df_tst_compact.is_fraud==1)
].yyhat_prob.hist()

sklearn.metrics.f1_score(
    df_tst_compact[df_tst_compact.amt<80].is_fraud,
    df_tst_compact[df_tst_compact.amt<80].yyhat_prob>0.5
)
0.5813528336380257
sklearn.metrics.recall_score(
    df_tst_compact[df_tst_compact.amt<80].is_fraud,
    df_tst_compact[df_tst_compact.amt<80].yyhat_prob>0.5
)
0.412987012987013
sklearn.metrics.precision_score(
    df_tst_compact[df_tst_compact.amt<80].is_fraud,
    df_tst_compact[df_tst_compact.amt<80].yyhat_prob>0.5
)
0.9814814814814815