Comparativa de performance de Modelo Apertura de MAIL: KPP vs GCP

Modelo old (kpp) entrenado en 202101 vs nuevo modelo (gcp) entrenado en 202206. Ambos con scoring para el 202209. Se compara el rendimiento en los meses posteriores.

import pandas as pd
import matplotlib.pyplot as plt
from sklearn import metrics
import seaborn as sns

tablon = pd.read_csv('TABLON2.csv',sep=',')

kpp = pd.read_csv('KPP2.csv',sep=';')

gcp = pd.read_csv('GCP2.csv',sep=';')

Modelo GCP - score 202209

fpr_gcp09, tpr_gcp09, thresholds_gcp09 = metrics.roc_curve(list(tablon['CONV_202209']),list(tablon['SC_MODELO_NEW']))
fpr_gcp10, tpr_gcp10, thresholds_gcp10 = metrics.roc_curve(list(tablon['CONV_202210']),list(tablon['SC_MODELO_NEW']))
fpr_gcp11, tpr_gcp11, thresholds_gcp11 = metrics.roc_curve(list(tablon['CONV_202211']),list(tablon['SC_MODELO_NEW']))

auc_gcp09 = metrics.roc_auc_score(list(tablon['CONV_202209']),list(tablon['SC_MODELO_NEW']))
auc_gcp10 = metrics.roc_auc_score(list(tablon['CONV_202210']),list(tablon['SC_MODELO_NEW']))
auc_gcp11 = metrics.roc_auc_score(list(tablon['CONV_202211']),list(tablon['SC_MODELO_NEW']))

fig, (ax1, ax2, ax3) = plt.subplots(1, 3,figsize=(15, 5))
fig.suptitle('Depreciación modelo GCP',fontsize=20)

ax1.plot(fpr_gcp09,tpr_gcp09,label='AUC = '+str(auc_gcp09))
ax2.plot(fpr_gcp10,tpr_gcp10,label='AUC = '+str(auc_gcp10))
ax3.plot(fpr_gcp11,tpr_gcp11,label='AUC = '+str(auc_gcp11))

ax1.set_title('202209')
ax1.set_ylabel('True positive Rate')
ax1.set_xlabel('False positive Rate')
ax2.set_title('202210')
ax2.set_ylabel('True positive Rate')
ax2.set_xlabel('False positive Rate')
ax3.set_title('202211')
ax3.set_ylabel('True positive Rate')
ax3.set_xlabel('False positive Rate')

ax1.legend(loc = "lower right")
ax2.legend(loc = "lower right")
ax3.legend(loc = "lower right")
plt.show()

Modelo KPP - score 202209

fpr_kpp09, tpr_kpp09, thresholds_kpp09 = metrics.roc_curve(list(tablon['CONV_202209']),list(tablon['SC_MODELO_OLD']))
fpr_kpp10, tpr_kpp10, thresholds_kpp10 = metrics.roc_curve(list(tablon['CONV_202210']),list(tablon['SC_MODELO_OLD']))
fpr_kpp11, tpr_kpp11, thresholds_kpp11 = metrics.roc_curve(list(tablon['CONV_202211']),list(tablon['SC_MODELO_OLD']))

auc_kpp09 = metrics.roc_auc_score(list(tablon['CONV_202209']),list(tablon['SC_MODELO_OLD']))
auc_kpp10 = metrics.roc_auc_score(list(tablon['CONV_202210']),list(tablon['SC_MODELO_OLD']))
auc_kpp11 = metrics.roc_auc_score(list(tablon['CONV_202211']),list(tablon['SC_MODELO_OLD']))

fig, (ax1, ax2, ax3) = plt.subplots(1, 3,figsize=(15, 5))
fig.suptitle('Depreciación modelo kpp',fontsize=20)

ax1.plot(fpr_kpp09,tpr_kpp09,label='AUC = '+str(auc_kpp09))
ax2.plot(fpr_kpp10,tpr_kpp10,label='AUC = '+str(auc_kpp10))
ax3.plot(fpr_kpp11,tpr_kpp11,label='AUC = '+str(auc_kpp11))


ax1.set_title('202209')
ax1.set_ylabel('True positive Rate')
ax1.set_xlabel('False positive Rate')
ax2.set_title('202210')
ax2.set_ylabel('True positive Rate')
ax2.set_xlabel('False positive Rate')
ax3.set_title('202211')
ax3.set_ylabel('True positive Rate')
ax3.set_xlabel('False positive Rate')

ax1.legend(loc = "lower right")
ax2.legend(loc = "lower right")
ax3.legend(loc = "lower right")
plt.show()

bines GCP

gcp = gcp.sort_values(by="MODELO_GCP_BIN")

gcp

<style scoped> .dataframe tbody tr th:only-of-type { vertical-align: middle; }

.dataframe tbody tr th {
    vertical-align: top;
}

.dataframe thead th {
    text-align: right;
}

</style>

	MODELO_GCP_BIN	Q_CLI	RR_202209	RR_202210	RR_202211
4	1	382105	0.049	0.076	0.062
8	2	390206	0.060	0.081	0.069
9	3	386155	0.064	0.077	0.070
5	4	386155	0.097	0.117	0.108
6	5	386118	0.175	0.182	0.164
7	6	386193	0.287	0.283	0.211
3	7	384594	0.504	0.488	0.382
2	8	387715	0.646	0.614	0.545
0	9	386151	0.679	0.648	0.565
1	10	386160	0.901	0.819	0.770

matplotlib.rc_file_defaults()
ax1 = sns.set_style(style="ticks")

fig, ax1= plt.subplots()

ax1 = sns.barplot(x=gcp["MODELO_GCP_BIN"],y = gcp["Q_CLI"], alpha=0.4, color="green")
ax1.set_title('Modelo GCP: Q_Clientes y RR según bines')
ax2 = ax1.twinx()
ax2 = sns.lineplot(data = gcp, x=gcp["MODELO_GCP_BIN"]-1, y = gcp["RR_202211"], label="rr - 202211")
plt.show()

bines KPP

kpp = kpp.sort_values(by="MODELO_OLD_BIN")

kpp

<style scoped> .dataframe tbody tr th:only-of-type { vertical-align: middle; }

.dataframe tbody tr th {
    vertical-align: top;
}

.dataframe thead th {
    text-align: right;
}

</style>

	MODELO_OLD_BIN	Q_CLI	RR_202209	RR_202210	RR_202211
3	1	1156359	0.025	0.092	0.081
7	2	338282	0.015	0.046	0.039
2	3	368829	0.172	0.194	0.162
6	4	450573	0.240	0.222	0.196
1	5	378993	0.384	0.355	0.306
4	6	369511	0.655	0.532	0.430
5	7	411902	0.879	0.761	0.641
0	8	387103	0.987	0.951	0.887

matplotlib.rc_file_defaults()
ax1 = sns.set_style(style="ticks")

fig, ax1= plt.subplots()

ax1 = sns.barplot(x=kpp["MODELO_OLD_BIN"],y = kpp["Q_CLI"], alpha=0.4, color="green")
ax1.set_title('Modelo KPP: Q_Clientes y RR según bines')
ax2 = ax1.twinx()
ax2 = sns.lineplot(data = kpp, x=kpp["MODELO_OLD_BIN"]-1, y = kpp["RR_202211"], label="rr - 202211")
plt.show()