TY - GEN
T1 - Transactional Compatible Representations for High Value Client Identification
T2 - 11th International Conference on Complex Networks, CompleNet 2020
AU - Unceta Mendieta, Irene
AU - Nin, J.
AU - Pujol, Oriol
N1 - Funding Information:
This work has been partially funded by the Spanish project TIN2016-74946-P (MINECO/FEDER, UE), and by AGAUR of the Generalitat de Catalunya through the Industrial PhD grant 2017-DI-25. We gratefully acknowledge the support of BBVA Data & Analytics for sponsoring the Industrial PhD.
Funding Information:
Acknowledgment. This work has been partially funded by the Spanish project TIN2016-74946-P (MINECO/FEDER, UE), and by AGAUR of the Generalitat de Catalunya through the Industrial PhD grant 2017-DI-25. We gratefully acknowledge the support of BBVA Data & Analytics for sponsoring the Industrial PhD.
Publisher Copyright:
© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020.
PY - 2020
Y1 - 2020
N2 - High value client identification is a crucial task to any company. In the banking industry, high value is not solely related to purchasing power, but also to an intensive use of financial services, such as card payments or bank wire transfers. This is why transactional data is a valuable source of information. In this work we propose a method to estimate the net worth of individuals for whom we lack any transactional data, either because they are non-clients or because they conduct their main activity elsewhere. We exploit the representation learned by a value prediction model trained over a signed graph of social financial relationships between BBVA clients to infer a transactional compatible representation of clients outside the graph. As a result, we obtain a new model that can predict value labels for both client and non-client data. Our results show an improvement in prediction accuracy over the previous baseline in a 2 million client database.
AB - High value client identification is a crucial task to any company. In the banking industry, high value is not solely related to purchasing power, but also to an intensive use of financial services, such as card payments or bank wire transfers. This is why transactional data is a valuable source of information. In this work we propose a method to estimate the net worth of individuals for whom we lack any transactional data, either because they are non-clients or because they conduct their main activity elsewhere. We exploit the representation learned by a value prediction model trained over a signed graph of social financial relationships between BBVA clients to infer a transactional compatible representation of clients outside the graph. As a result, we obtain a new model that can predict value labels for both client and non-client data. Our results show an improvement in prediction accuracy over the previous baseline in a 2 million client database.
KW - Financial network
KW - Graph embeddings
KW - Value prediction
UR - http://www.scopus.com/inward/record.url?scp=85081238768&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-40943-2_28
DO - 10.1007/978-3-030-40943-2_28
M3 - Conference contribution
AN - SCOPUS:85081238768
SN - 9783030409425
T3 - Springer Proceedings in Complexity
SP - 334
EP - 345
BT - Complex Networks XI - Proceedings of the 11th Conference on Complex Networks, CompleNet 2020
A2 - Barbosa, Hugo
A2 - Menezes, Ronaldo
A2 - Gomez-Gardenes, Jesus
A2 - Gonçalves, Bruno
A2 - Mangioni, Giuseppe
A2 - Oliveira, Marcos
PB - Springer
Y2 - 31 March 2020 through 3 April 2020
ER -