
import pandas as pd 
import matplotlib.pyplot as plt 
import numpy as np


data= pd.read_csv("train.csv")
data_test= pd.read_csv("test.csv")


#data["FECHA"]


data_test["Dist_max_INTER"]#.columns
data_test.columns.values.shape

##.dtypes
data_test.columns.values#.columns

array(['id', 'FRAUDE', 'VALOR', 'HORA_AUX', 'Dist_max_COL',
       'Dist_max_INTER', 'Canal1', 'FECHA_FRAUDE', 'COD_PAIS', 'CANAL',
       'FECHA', 'DIASEM', 'DIAMES', 'FECHA_VIN', 'OFICINA_VIN', 'SEXO',
       'SEGMENTO', 'EDAD', 'INGRESOS', 'EGRESOS', 'NROPAISES',
       'Dist_Sum_INTER', 'Dist_Mean_INTER', 'Dist_Max_INTER',
       'NROCIUDADES', 'Dist_Sum_NAL', 'Dist_Mean_NAL', 'Dist_HOY',
       'Dist_sum_NAL', 'Dist_mean_NAL', 'Dist_sum_INTER',
       'Dist_mean_INTER'], dtype=object)


#data[data["id"]==98523068]
data.columns.values

array(['id', 'FRAUDE', 'VALOR', 'HORA_AUX', 'Dist_max_NAL', 'Canal1',
       'FECHA', 'COD_PAIS', 'CANAL', 'DIASEM', 'DIAMES', 'FECHA_VIN',
       'OFICINA_VIN', 'SEXO', 'SEGMENTO', 'EDAD', 'INGRESOS', 'EGRESOS',
       'NROPAISES', 'Dist_Sum_INTER', 'Dist_Mean_INTER', 'Dist_Max_INTER',
       'NROCIUDADES', 'Dist_Mean_NAL', 'Dist_HOY', 'Dist_sum_NAL'],
      dtype=object)


# Hay desbalanceo de datos, se puede aplicar el metodo de sobremuestreo con SMOTE
data["FRAUDE"].value_counts()

0    2234
1     731
Name: FRAUDE, dtype: int64


731/(2234+731)

0.24654300168634063


# Obtener la intersección de columnas
common_columns = np.intersect1d(data_test.columns.values,data.columns.values  )

#print("Columnas comunes en ambos conjuntos de datos:")
#print(common_columns)

# Obtener la intersección de columnas
common_columns = np.intersect1d(data.columns.values, data_test.columns.values)

# Ordenar las columnas comunes en el mismo orden que el primer array
common_columns_sorted = data_test.columns.values[np.isin(data_test.columns.values, common_columns)]

print("Columnas comunes en ambos conjuntos de datos en el mismo orden que el primer array:")
print(common_columns_sorted)

Columnas comunes en ambos conjuntos de datos en el mismo orden que el primer array:
['id' 'FRAUDE' 'VALOR' 'HORA_AUX' 'Canal1' 'COD_PAIS' 'CANAL' 'FECHA'
 'DIASEM' 'DIAMES' 'FECHA_VIN' 'OFICINA_VIN' 'SEXO' 'SEGMENTO' 'EDAD'
 'INGRESOS' 'EGRESOS' 'NROPAISES' 'Dist_Sum_INTER' 'Dist_Mean_INTER'
 'Dist_Max_INTER' 'NROCIUDADES' 'Dist_Mean_NAL' 'Dist_HOY' 'Dist_sum_NAL']


#data.drop([Dist_Sum_INTER,Dist_Mean_INTER ])
# Seleccionar solo las columnas comunes
data_train = data[common_columns_sorted]


data_train.isnull().sum()

id                    0
FRAUDE                0
VALOR                 0
HORA_AUX              0
Canal1                0
COD_PAIS              0
CANAL                 0
FECHA                 0
DIASEM                0
DIAMES                0
FECHA_VIN            24
OFICINA_VIN          24
SEXO                 55
SEGMENTO             24
EDAD                 24
INGRESOS             24
EGRESOS              24
NROPAISES             0
Dist_Sum_INTER     1547
Dist_Mean_INTER    1547
Dist_Max_INTER     1547
NROCIUDADES           0
Dist_Mean_NAL       457
Dist_HOY              0
Dist_sum_NAL          0
dtype: int64


data_train


data_train.isnull().sum()
data_train["SEGMENTO"].value_counts()

Personal Plus    1527
Preferencial      958
Personal          174
Emprendedor       159
PYME              119
Empresarial         4
Name: SEGMENTO, dtype: int64


# Contar el número de transacciones con y sin fraude
fraude_counts = data['FRAUDE'].value_counts()

# Crear un gráfico de barras
plt.figure(figsize=(8, 6))
fraude_counts.plot(kind='bar', color=['blue', 'red'])
plt.title('Número de Transacciones con y sin Fraude')
plt.xlabel('Fraude')
plt.ylabel('Número de Transacciones')
plt.xticks(ticks=[0, 1], labels=['Sin Fraude', 'Con Fraude'], rotation=0)
plt.show()


# Contar el número de transacciones por día de la semana (DIASEM)
transactions_by_day = data_train.groupby('DIASEM')['id'].count()

# Contar el número de transacciones fraudulentas por día de la semana (DIASEM)
fraud_transactions_by_day = data_train[data_train['FRAUDE'] == 1].groupby('DIASEM')['id'].count()

# Contar el número de transacciones por día del mes (DIAMES)
transactions_by_date = data_train.groupby('DIAMES')['id'].count()

# Contar el número de transacciones fraudulentas por día del mes (DIAMES)
fraud_transactions_by_date = data_train[data_train['FRAUDE'] == 1].groupby('DIAMES')['id'].count()

# Crear figura y ejes
fig, axs = plt.subplots(2, 1, figsize=(10, 10))

# Gráfico de barras apiladas para transacciones por día de la semana

#transactions_by_day.plot(kind='bar', color='blue', ax=axs[0], label='Transacciones')
fraud_transactions_by_day.plot(kind='bar', color='red', ax=axs[0], label='Fraudulentas')
axs[0].set_title('Transacciones por Día de la Semana')
axs[0].set_xlabel('Día de la Semana')
axs[0].set_ylabel('Número de Transacciones')
axs[0].legend()

# Gráfico de barras apiladas para transacciones por día del mes
fraud_transactions_by_date.plot(kind='bar', color='red', ax=axs[1], label='Fraudulentas')
axs[1].set_title('Transacciones por Día del Mes')
axs[1].set_xlabel('Día del Mes')
axs[1].set_ylabel('Número de Transacciones')
axs[1].legend()

plt.tight_layout()
plt.show()


# Descartar las columnas mencionadas en una sola línea con inplace=True
data_train.drop(columns=['Dist_Sum_INTER', 'Dist_Mean_INTER', 'Dist_Max_INTER',"COD_PAIS","id"
                         ,"FECHA_VIN","OFICINA_VIN","CANAL","FECHA"], inplace=True)

/tmp/ipykernel_147672/3244588327.py:2: SettingWithCopyWarning: 
A value is trying to be set on a copy of a slice from a DataFrame

See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy
  data_train.drop(columns=['Dist_Sum_INTER', 'Dist_Mean_INTER', 'Dist_Max_INTER',"COD_PAIS","id"


#print(data_train["CANAL"].value_counts())
print(data_train["Canal1"].value_counts())
data_train.isnull().sum()

POS        2329
ATM_INT     636
Name: Canal1, dtype: int64

FRAUDE             0
VALOR              0
HORA_AUX           0
Canal1             0
DIASEM             0
DIAMES             0
SEXO              55
SEGMENTO          24
EDAD              24
INGRESOS          24
EGRESOS           24
NROPAISES          0
NROCIUDADES        0
Dist_Mean_NAL    457
Dist_HOY           0
Dist_sum_NAL       0
dtype: int64


# Codificación one-hot para las variables categóricas
data_encoded = pd.get_dummies(data_train, columns=['Canal1', 'SEXO', 'SEGMENTO'])

# Ver las primeras filas del conjunto de datos codificado


#Datos que se utilizaran para el modelo
data_encoded


from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import classification_report
from imblearn.over_sampling import SMOTE
from sklearn.feature_selection import SelectFromModel


# Realizar selección de características utilizando la importancia de características
#X = data_encoded.drop(columns=['FRAUDE'])
#y = data_encoded['FRAUDE']
#clf = RandomForestClassifier(n_estimators=100, random_state=42)
#clf.fit(X, y)
#feature_selector = SelectFromModel(clf, prefit=True)
#X_selected = feature_selector.transform(X)


from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import classification_report
from imblearn.over_sampling import SMOTE
from sklearn.feature_selection import SelectFromModel
from sklearn.impute import SimpleImputer
# Realizar selección de características utilizando la importancia de características
X = data_encoded.drop(columns=['FRAUDE'])
y = data_encoded['FRAUDE']
# Imputar valores faltantes
imputer = SimpleImputer(strategy='mean')
X_imputed = imputer.fit_transform(X)

# Realizar selección de características utilizando la importancia de características


lf = RandomForestClassifier(n_estimators=100, random_state=42)
lf.fit(X_imputed, y)
feature_selector = SelectFromModel(lf, prefit=True)
X_selected = feature_selector.transform(X_imputed)


# Dividir los datos en conjuntos de entrenamiento y prueba
X_train, X_test, y_train, y_test = train_test_split(X_selected, y, test_size=0.2, random_state=42)

# Aplicar SMOTE al conjunto de entrenamiento (opcional)
smote = SMOTE(random_state=42)
X_train_smote, y_train_smote = smote.fit_resample(X_train, y_train)


# Entrenar el modelo de detección de fraude
model = RandomForestClassifier(n_estimators=100, random_state=42)
model.fit(X_train, y_train)

# Evaluar el modelo en el conjunto de prueba
y_pred = model.predict(X_test)
print(classification_report(y_test, y_pred))

              precision    recall  f1-score   support

           0       0.97      0.98      0.97       456
           1       0.92      0.91      0.91       137

    accuracy                           0.96       593
   macro avg       0.95      0.94      0.94       593
weighted avg       0.96      0.96      0.96       593


X_train.shape
X_test.shape
X_train_smote.shape

(3556, 10)


from sklearn.feature_selection import SelectFromModel
from sklearn.ensemble import RandomForestClassifier
from sklearn.linear_model import LogisticRegression
import xgboost as xgb
import lightgbm as lgb
# Modelos de clasificación
classifiers = {
    "Random Forest": RandomForestClassifier(n_estimators=100, random_state=42),
    "Logistic Regression": LogisticRegression(random_state=42),
    "XGBoost": xgb.XGBClassifier(random_state=42),
    "LightGBM": lgb.LGBMClassifier(random_state=42)
}

# Entrenamiento y evaluación de modelos
for name, classifier in classifiers.items():
    print(f"Entrenando modelo {name}")
    if name == "XGBoost" or name == "LightGBM":
        classifier.fit(X_train_smote, y_train_smote, eval_metric='logloss')
    else:
        classifier.fit(X_train_smote, y_train_smote)
    
    # Predicción y evaluación en conjunto de prueba
    y_pred = classifier.predict(X_test)
    print(f"Resultados para {name}:")
    print(classification_report(y_test, y_pred))
    print("\n")

Entrenando modelo Random Forest
Resultados para Random Forest:
              precision    recall  f1-score   support

           0       0.97      0.96      0.96       456
           1       0.86      0.90      0.88       137

    accuracy                           0.94       593
   macro avg       0.91      0.93      0.92       593
weighted avg       0.94      0.94      0.94       593


Entrenando modelo Logistic Regression
Resultados para Logistic Regression:
              precision    recall  f1-score   support

           0       0.87      0.61      0.72       456
           1       0.35      0.69      0.46       137

    accuracy                           0.63       593
   macro avg       0.61      0.65      0.59       593
weighted avg       0.75      0.63      0.66       593


Entrenando modelo XGBoost

/home/jordan/.local/lib/python3.10/site-packages/xgboost/sklearn.py:835: UserWarning: `eval_metric` in `fit` method is deprecated for better compatibility with scikit-learn, use `eval_metric` in constructor or`set_params` instead.
  warnings.warn(

Resultados para XGBoost:
              precision    recall  f1-score   support

           0       0.97      0.97      0.97       456
           1       0.91      0.90      0.90       137

    accuracy                           0.96       593
   macro avg       0.94      0.94      0.94       593
weighted avg       0.96      0.96      0.96       593


Entrenando modelo LightGBM
Resultados para LightGBM:
              precision    recall  f1-score   support

           0       0.97      0.96      0.97       456
           1       0.89      0.91      0.90       137

    accuracy                           0.95       593
   macro avg       0.93      0.94      0.93       593
weighted avg       0.95      0.95      0.95       593


import matplotlib.pyplot as plt
from sklearn.metrics import roc_auc_score, roc_curve

# Función para trazar la curva ROC-AUC
def plot_roc_curve(y_true, y_pred_probs, model_name):
    auc = roc_auc_score(y_true, y_pred_probs)
    fpr, tpr, _ = roc_curve(y_true, y_pred_probs)
    plt.plot(fpr, tpr, label=f'{model_name} (AUC = {auc:.2f})')
    plt.xlabel('False Positive Rate')
    plt.ylabel('True Positive Rate')
    plt.title('ROC Curve')
    plt.legend()
    plt.grid(True)

# Entrenar y evaluar modelos
for name, classifier in classifiers.items():
    print(f"Entrenando modelo {name}")
    if name == "XGBoost" or name == "LightGBM":
        classifier.fit(X_train_smote, y_train_smote, eval_metric='logloss')
        y_pred_probs = classifier.predict_proba(X_test)[:, 1]
    else:
        classifier.fit(X_train_smote, y_train_smote)
        y_pred_probs = classifier.predict_proba(X_test)[:, 1]
    
    # Trazar la curva ROC-AUC
    plot_roc_curve(y_test, y_pred_probs, name)

# Mostrar la curva ROC-AUC para todos los modelos
plt.plot([0, 1], [0, 1], 'k--', label='Random Guess')
plt.xlim([0.0, 1.0])
plt.ylim([0.0, 1.05])
plt.legend(loc='lower right')
plt.show()

Entrenando modelo Random Forest
Entrenando modelo Logistic Regression
Entrenando modelo XGBoost

/home/jordan/.local/lib/python3.10/site-packages/xgboost/sklearn.py:835: UserWarning: `eval_metric` in `fit` method is deprecated for better compatibility with scikit-learn, use `eval_metric` in constructor or`set_params` instead.
  warnings.warn(

Entrenando modelo LightGBM


data_test1= data_test[common_columns_sorted]


# Descartar las columnas mencionadas en una sola línea con inplace=True
data_test1.drop(columns=['Dist_Sum_INTER', 'Dist_Mean_INTER', 'Dist_Max_INTER',"COD_PAIS","id"
                         ,"FECHA_VIN","OFICINA_VIN","CANAL","FECHA"], inplace=True)

/tmp/ipykernel_147672/1368707283.py:2: SettingWithCopyWarning: 
A value is trying to be set on a copy of a slice from a DataFrame

See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy
  data_test1.drop(columns=['Dist_Sum_INTER', 'Dist_Mean_INTER', 'Dist_Max_INTER',"COD_PAIS","id"


data_test1


# Codificación one-hot para las variables categóricas
data_encoded = pd.get_dummies(data_test1, columns=['Canal1', 'SEXO', 'SEGMENTO'])


data_encoded.isnull().sum()

FRAUDE                    100
VALOR                       0
HORA_AUX                    0
DIASEM                      0
DIAMES                      0
EDAD                        0
INGRESOS                    0
EGRESOS                     0
NROPAISES                   0
NROCIUDADES                 0
Dist_Mean_NAL              21
Dist_HOY                    0
Dist_sum_NAL                0
Canal1_ATM_INT              0
Canal1_POS                  0
SEXO_F                      0
SEXO_M                      0
SEGMENTO_Emprendedor        0
SEGMENTO_PYME               0
SEGMENTO_Personal           0
SEGMENTO_Personal Plus      0
SEGMENTO_Preferencial       0
dtype: int64


from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import classification_report
from imblearn.over_sampling import SMOTE
from sklearn.feature_selection import SelectFromModel
from sklearn.impute import SimpleImputer


data_encoded1 = data_encoded[['VALOR', 'HORA_AUX','DIAMES', 'EDAD', 'INGRESOS', 'EGRESOS',
       'Dist_Mean_NAL', 'Dist_sum_NAL', 'Canal1_ATM_INT', 'Canal1_POS']]


X_= data_encoded1


from sklearn.impute import KNNImputer

imputeKNN= KNNImputer(n_neighbors=2)

#SimpleImputer(strategy='mean')
X_imputed = imputeKNN.fit_transform(X_)


X_imputed.shape

(100, 10)


# Evaluar el modelo en el conjunto de prueba
y_pred = model.predict(X_imputed)


#PREDICCIONES
y_pred

array([0, 0, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 1,
       1, 0, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1,
       1, 0, 1, 1, 1, 1, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1,
       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0])


data_test


data_test.loc[:,"FRAUDE"] = list(y_pred)
#df.loc[:, 'Column1']

/tmp/ipykernel_147672/3603763774.py:1: FutureWarning: In a future version, `df.iloc[:, i] = newvals` will attempt to set the values inplace instead of always setting a new array. To retain the old behavior, use either `df[df.columns[i]] = newvals` or, if columns are non-unique, `df.isetitem(i, newvals)`
  data_test.loc[:,"FRAUDE"] = list(y_pred)


#data_test1["Fraude"].shape


data_test


data_test


data_test.to_csv("test_evaluado.csv")

(2965, 22)


selected_indices

array([ 0,  1,  3,  4,  5,  6,  9, 11, 12, 13])


import numpy as np

# Obtener índices de las características seleccionadas
selected_indices = np.where(feature_selector.get_support())[0]

# Obtener nombres de las características seleccionadas
selected_features = X.columns[selected_indices]

print("Características seleccionadas:")
print(selected_features)
selected_features.shape

Características seleccionadas:
Index(['VALOR', 'HORA_AUX', 'DIAMES', 'EDAD', 'INGRESOS', 'EGRESOS',
       'Dist_Mean_NAL', 'Dist_sum_NAL', 'Canal1_ATM_INT', 'Canal1_POS'],
      dtype='object')

(10,)


import matplotlib.pyplot as plt

# Obtener la importancia de las características del clasificador de Bosques Aleatorios
feature_importances = lf.feature_importances_

# Crear un DataFrame para visualizar la importancia de las características
feature_importance_df = pd.DataFrame({'Feature': X.columns, 'F-Score': feature_importances})

# Ordenar las características por su importancia de acuerdo al F-score
feature_importance_df.sort_values(by='F-Score', ascending=False, inplace=True)

# Visualizar la importancia de las características en un gráfico de barras
plt.figure(figsize=(10, 6))
plt.barh(feature_importance_df['Feature'], feature_importance_df['F-Score'], color='skyblue')
plt.xlabel('F-Score')
plt.ylabel('Característica')
plt.title('Importancia de las Características (F-Score)')
plt.gca().invert_yaxis()  # Invertir el eje y para mostrar la característica más importante en la parte superior
plt.show()

	id	FRAUDE	VALOR	HORA_AUX	Canal1	COD_PAIS	CANAL	FECHA	DIASEM	DIAMES	...	INGRESOS	EGRESOS	NROPAISES	Dist_Sum_INTER	Dist_Mean_INTER	Dist_Max_INTER	NROCIUDADES	Dist_Mean_NAL	Dist_HOY	Dist_sum_NAL
0	9000000001	1	0.00	13	ATM_INT	US	ATM_INT	20150501	5	1	...	1200000.0	1200000.0	1	NaN	NaN	NaN	6	474.94	4552.41	5224.36
1	9000000002	1	0.00	17	ATM_INT	US	ATM_INT	20150515	5	15	...	5643700.0	500000.0	1	NaN	NaN	NaN	5	289.99	4552.41	2029.90
2	9000000003	1	0.00	13	ATM_INT	US	ATM_INT	20150501	5	1	...	1200000.0	1200000.0	1	NaN	NaN	NaN	6	474.94	4552.41	5224.36
3	9000000004	1	0.00	13	ATM_INT	US	ATM_INT	20150501	5	1	...	1200000.0	1200000.0	1	NaN	NaN	NaN	6	474.94	4552.41	5224.36
4	9000000005	1	0.00	0	ATM_INT	CR	ATM_INT	20150510	0	10	...	0.0	0.0	1	NaN	NaN	NaN	1	NaN	1482.35	1.00
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
2960	622529101	1	993430.04	19	POS	US	POS	20150519	2	19	...	103918285.0	95475378.0	4	8944.83	2236.21	3646.67	4	96.86	4552.41	484.30
2961	2043206272	0	9957.05	10	POS	US	POS	20150524	0	24	...	23625000.0	5000000.0	3	27648.32	3949.76	4552.41	11	82.67	4552.41	2810.75
2962	2943206272	0	9957.05	10	POS	US	POS	20150524	0	24	...	23625000.0	5000000.0	3	27648.32	3949.76	4552.41	11	82.67	4552.41	2810.75
2963	3136302872	0	996191.64	15	POS	US	MCI	20150513	3	13	...	56666000.0	37600750.0	1	NaN	NaN	NaN	3	219.46	4552.41	1316.79
2964	1953178702	1	999276.60	16	ATM_INT	CR	ATM_INT	20150520	3	20	...	12853000.0	6156000.0	1	NaN	NaN	NaN	1	NaN	1482.35	1.00

	FRAUDE	VALOR	HORA_AUX	DIASEM	DIAMES	EDAD	INGRESOS	EGRESOS	NROPAISES	NROCIUDADES	...	Canal1_ATM_INT	Canal1_POS	SEXO_F	SEXO_M	SEGMENTO_Emprendedor	SEGMENTO_Empresarial	SEGMENTO_PYME	SEGMENTO_Personal	SEGMENTO_Personal Plus	SEGMENTO_Preferencial
0	1	0.00	13	5	1	29.0	1200000.0	1200000.0	1	6	...	1	0	0	1	0	0	0	0	1	0
1	1	0.00	17	5	15	29.0	5643700.0	500000.0	1	5	...	1	0	0	1	0	0	0	0	1	0
2	1	0.00	13	5	1	29.0	1200000.0	1200000.0	1	6	...	1	0	0	1	0	0	0	0	1	0
3	1	0.00	13	5	1	29.0	1200000.0	1200000.0	1	6	...	1	0	0	1	0	0	0	0	1	0
4	1	0.00	0	0	10	25.0	0.0	0.0	1	1	...	1	0	0	1	0	0	0	1	0	0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
2960	1	993430.04	19	2	19	48.0	103918285.0	95475378.0	4	4	...	0	1	1	0	0	0	0	0	0	1
2961	0	9957.05	10	0	24	35.0	23625000.0	5000000.0	3	11	...	0	1	1	0	0	0	0	0	0	1
2962	0	9957.05	10	0	24	35.0	23625000.0	5000000.0	3	11	...	0	1	1	0	0	0	0	0	0	1
2963	0	996191.64	15	3	13	34.0	56666000.0	37600750.0	1	3	...	0	1	1	0	0	0	1	0	0	0
2964	1	999276.60	16	3	20	29.0	12853000.0	6156000.0	1	1	...	1	0	1	0	0	0	0	0	1	0

	FRAUDE	VALOR	HORA_AUX	Canal1	DIASEM	DIAMES	SEXO	SEGMENTO	EDAD	INGRESOS	EGRESOS	NROPAISES	NROCIUDADES	Dist_Mean_NAL	Dist_HOY	Dist_sum_NAL
0	NaN	42230.09	18	POS	5	15	F	Personal Plus	46	20000000	10000000	1	1	NaN	4552.41	1.00
1	NaN	143202.65	20	POS	3	6	F	Preferencial	56	11000000	4500000	3	2	614.04	4552.41	1228.07
2	NaN	243591.25	2	ATM_INT	0	17	F	Personal Plus	33	9000000	4000000	3	7	138.88	5083.41	1944.35
3	NaN	238267.40	20	ATM_INT	5	8	F	Personal Plus	53	2300000	500000	1	1	NaN	904.81	1.00
4	NaN	490403.58	13	ATM_INT	5	1	M	Personal	0	0	0	1	1	NaN	4552.41	1.00
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
95	NaN	531534.03	13	POS	5	1	M	Preferencial	51	19327667	3000000	1	4	56.29	4552.41	844.35
96	NaN	52035.08	11	POS	0	3	M	Preferencial	40	35500000	2000000	2	3	25.25	971.23	151.52
97	NaN	18309.04	23	POS	5	15	M	Personal Plus	43	3100000	2000000	1	2	61.45	4552.41	122.90
98	NaN	496906.75	20	ATM_INT	6	16	M	Personal Plus	40	24000000	10500000	1	3	453.23	4552.41	1812.93
99	NaN	192825.50	20	POS	5	15	M	Preferencial	36	47736000	3000000	2	9	113.45	4552.41	5218.81

	id	FRAUDE	VALOR	HORA_AUX	Dist_max_COL	Dist_max_INTER	Canal1	FECHA_FRAUDE	COD_PAIS	CANAL	...	Dist_Mean_INTER	Dist_Max_INTER	NROCIUDADES	Dist_Sum_NAL	Dist_Mean_NAL	Dist_HOY	Dist_sum_NAL	Dist_mean_NAL	Dist_sum_INTER	Dist_mean_INTER
0	98523068	NaN	42230.09	18	1.00	1.00	POS	20150515	US	POS	...	NaN	NaN	1	NaN	NaN	4552.41	1.00	1.00	1.00	1.00
1	300237898	NaN	143202.65	20	614.04	7632.97	POS	20150506	US	MCI	...	6092.69	7632.97	2	1228.07	614.04	4552.41	1228.07	614.04	24370.75	6092.69
2	943273308	NaN	243591.25	2	286.84	2443.14	ATM_INT	20150517	EC	ATM_INT	...	1743.52	2443.14	7	1944.35	138.88	5083.41	1944.35	138.88	6974.09	1743.52
3	951645809	NaN	238267.40	20	1.00	1.00	ATM_INT	20150508	EC	ATM_INT	...	NaN	NaN	1	NaN	NaN	904.81	1.00	1.00	1.00	1.00
4	963797516	NaN	490403.58	13	1.00	1.00	ATM_INT	20150501	US	ATM_INT	...	NaN	NaN	1	NaN	NaN	4552.41	1.00	1.00	1.00	1.00
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
95	9970518152	NaN	531534.03	13	340.09	1.00	POS	20150501	US	POS	...	NaN	NaN	4	844.35	56.29	4552.41	844.35	56.29	1.00	1.00
96	9971748725	NaN	52035.08	11	28.59	4552.41	POS	20150503	AW	POS	...	4552.41	4552.41	3	151.52	25.25	971.23	151.52	25.25	9104.82	4552.41
97	9979565282	NaN	18309.04	23	61.45	1.00	POS	20150515	US	POS	...	NaN	NaN	2	122.90	61.45	4552.41	122.90	61.45	1.00	1.00
98	9979718478	NaN	496906.75	20	733.11	1.00	ATM_INT	20150516	US	ATM_INT	...	NaN	NaN	3	1812.93	453.23	4552.41	1812.93	453.23	1.00	1.00
99	9998668320	NaN	192825.50	20	337.29	904.81	POS	20150515	US	MCI	...	904.81	904.81	9	5218.81	113.45	4552.41	5218.81	113.45	1809.62	904.81

	id	FRAUDE	VALOR	HORA_AUX	Dist_max_COL	Dist_max_INTER	Canal1	FECHA_FRAUDE	COD_PAIS	CANAL	...	Dist_Mean_INTER	Dist_Max_INTER	NROCIUDADES	Dist_Sum_NAL	Dist_Mean_NAL	Dist_HOY	Dist_sum_NAL	Dist_mean_NAL	Dist_sum_INTER	Dist_mean_INTER
0	98523068	0	42230.09	18	1.00	1.00	POS	20150515	US	POS	...	NaN	NaN	1	NaN	NaN	4552.41	1.00	1.00	1.00	1.00
1	300237898	0	143202.65	20	614.04	7632.97	POS	20150506	US	MCI	...	6092.69	7632.97	2	1228.07	614.04	4552.41	1228.07	614.04	24370.75	6092.69
2	943273308	1	243591.25	2	286.84	2443.14	ATM_INT	20150517	EC	ATM_INT	...	1743.52	2443.14	7	1944.35	138.88	5083.41	1944.35	138.88	6974.09	1743.52
3	951645809	1	238267.40	20	1.00	1.00	ATM_INT	20150508	EC	ATM_INT	...	NaN	NaN	1	NaN	NaN	904.81	1.00	1.00	1.00	1.00
4	963797516	1	490403.58	13	1.00	1.00	ATM_INT	20150501	US	ATM_INT	...	NaN	NaN	1	NaN	NaN	4552.41	1.00	1.00	1.00	1.00
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
95	9970518152	0	531534.03	13	340.09	1.00	POS	20150501	US	POS	...	NaN	NaN	4	844.35	56.29	4552.41	844.35	56.29	1.00	1.00
96	9971748725	0	52035.08	11	28.59	4552.41	POS	20150503	AW	POS	...	4552.41	4552.41	3	151.52	25.25	971.23	151.52	25.25	9104.82	4552.41
97	9979565282	0	18309.04	23	61.45	1.00	POS	20150515	US	POS	...	NaN	NaN	2	122.90	61.45	4552.41	122.90	61.45	1.00	1.00
98	9979718478	0	496906.75	20	733.11	1.00	ATM_INT	20150516	US	ATM_INT	...	NaN	NaN	3	1812.93	453.23	4552.41	1812.93	453.23	1.00	1.00
99	9998668320	0	192825.50	20	337.29	904.81	POS	20150515	US	MCI	...	904.81	904.81	9	5218.81	113.45	4552.41	5218.81	113.45	1809.62	904.81

Modelo antifraude¶

Desarrollo de un modelo que permite, a partir de los datos en el archivo train.csv predecir cuál será el valor de la variable FRAUDE para una transacción cualquiera. Donde la variable FRAUDE es 1 si la transacción constituyó un fraude o 0 si fue una transacción legítima¶

Aplicación al conjunto de datos data_test¶

En ningun momento se hace una reorganización de los datos utilizados para realizar la predicción por lo tanto están ordenados tal como estaban inicialmente¶

## Anexo de variables escogidas utilizando feauture importance¶