Modelos predictivos para detectar el fraude en el consumo de agua
Loading...
Official URL
Full text at PDC
Publication date
2020
Defense date
07/2020
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citation
Abstract
El presente estudio abordará el análisis de posibles fraudes cometidos en el seno de una compañía de servicios de agua en España. Para ello, se hará uso de diferentes datos obtenidos de los clientes de una comunidad autónoma de nuestro país, a través de los cuales se tratará de obtener el mejor modelo de predicción de fraude con la finalidad de evitar que estos hechos se sigan cometiendo.
La metodología para llevar a cabo este trabajo se basará en diferentes técnicas estadísticas, comenzando por la depuración de nuestra base de datos y avanzando hasta la Regresión Logística, Redes Neuronales o Bosques Aleatorios. Estas tareas ayudarán a obtener un modelo robusto y fiable para la detección de irregularidades en base a las características y comportamientos de estos clientes.
El objetivo final será el de detectar todos los usuarios fraudulentos de la empresa mediante el modelo obtenido para que la misma pueda visitar a todos los clientes que indique el mismo.
This study will address the analysis of possible fraud committed within a water services company in Spain. To do this, different customer data will be use from an autonomous community in our country, through which the study will try to obtain the best fraud prediction model with the resolution of preventing these acts from continuing to be committed. The methodology will be based on different statistical techniques, starting with the debugging of our database and progressing to logistic regression, neural networks or random forests. These tasks help us to obtain a robust and reliable model for detecting irregularities based on the characteristics and behaviour of these clients. The final objective will be to detect all the fraudulent users of the company through our model so the company could visit all the clients indicated by the model.
This study will address the analysis of possible fraud committed within a water services company in Spain. To do this, different customer data will be use from an autonomous community in our country, through which the study will try to obtain the best fraud prediction model with the resolution of preventing these acts from continuing to be committed. The methodology will be based on different statistical techniques, starting with the debugging of our database and progressing to logistic regression, neural networks or random forests. These tasks help us to obtain a robust and reliable model for detecting irregularities based on the characteristics and behaviour of these clients. The final objective will be to detect all the fraudulent users of the company through our model so the company could visit all the clients indicated by the model.