TY - GEN
T1 - Leveraging Unstructured Data to Improve Customer Engagement and Revenue in Financial Institutions
T2 - 2023 IEEE International Conference on Computer, Information, and Telecommunication Systems, CITS 2023
AU - Jain, Shubham
AU - Fallon, Enda
N1 - Publisher Copyright:
© 2023 IEEE.
PY - 2023
Y1 - 2023
N2 - Personalized customer transaction insights have become increasingly important for financial institutions to increase customer engagement and revenue. In this paper, we propose a novel methodology that combines deep learning and reinforcement learning techniques to extract features from customer demographics and capture temporal dynamics of customer transactions to provide personalized credit card recommendations. Our methodology is designed to learn a policy that maximizes the expected reward for providing personalized recommendations to customers. To gain additional insights about customer preferences and sentiments, we utilize both structured and unstructured data, including the transaction descriptions. To process the unstructured transaction descriptions, we use natural language processing techniques to extract relevant features and insights. Incorporating unstructured data into the proposed methodology can improve the accuracy and effectiveness of the customer transaction insights, and can provide a more comprehensive understanding of customer preferences and behaviors. We test our proposed methodology on the publicly available IEEE-CIS Fraud Detection dataset, which contains over 590,000 transactions with information such as transaction amount, transaction date and time, and transaction description. For the test set, we achieve an AUC-ROC score of 0.9485, outperforming existing techniques for customer transaction insights. Our findings show that delivering customised credit card suggestions to customers has the potential to increase revenue for financial institutions. Overall, the use of deep learning and reinforcement learning techniques, as well as structured and unstructured data, can provide a more thorough understanding of consumer transaction data, allowing for more accurate and effective personalized suggestions. The use of unstructured data is a crucial component of our technique since it can give financial institutions with new insights and value. It is critical to appropriately anonymize and secure the customers whose data is used in the study.
AB - Personalized customer transaction insights have become increasingly important for financial institutions to increase customer engagement and revenue. In this paper, we propose a novel methodology that combines deep learning and reinforcement learning techniques to extract features from customer demographics and capture temporal dynamics of customer transactions to provide personalized credit card recommendations. Our methodology is designed to learn a policy that maximizes the expected reward for providing personalized recommendations to customers. To gain additional insights about customer preferences and sentiments, we utilize both structured and unstructured data, including the transaction descriptions. To process the unstructured transaction descriptions, we use natural language processing techniques to extract relevant features and insights. Incorporating unstructured data into the proposed methodology can improve the accuracy and effectiveness of the customer transaction insights, and can provide a more comprehensive understanding of customer preferences and behaviors. We test our proposed methodology on the publicly available IEEE-CIS Fraud Detection dataset, which contains over 590,000 transactions with information such as transaction amount, transaction date and time, and transaction description. For the test set, we achieve an AUC-ROC score of 0.9485, outperforming existing techniques for customer transaction insights. Our findings show that delivering customised credit card suggestions to customers has the potential to increase revenue for financial institutions. Overall, the use of deep learning and reinforcement learning techniques, as well as structured and unstructured data, can provide a more thorough understanding of consumer transaction data, allowing for more accurate and effective personalized suggestions. The use of unstructured data is a crucial component of our technique since it can give financial institutions with new insights and value. It is critical to appropriately anonymize and secure the customers whose data is used in the study.
KW - Deep Learning
KW - Reinforcement Learning
KW - Semantic Data Extraction
UR - http://www.scopus.com/inward/record.url?scp=85167869763&partnerID=8YFLogxK
U2 - 10.1109/CITS58301.2023.10188745
DO - 10.1109/CITS58301.2023.10188745
M3 - Conference contribution
AN - SCOPUS:85167869763
SN - 9798350336092
T3 - Proceedings of the 2023 IEEE International Conference on Computer, Information, and Telecommunication Systems, CITS 2023
BT - Proceedings of the 2023 IEEE International Conference on Computer, Information, and Telecommunication Systems, CITS 2023
A2 - Obaidat, Mohammad S.
A2 - Obaidat, Mohammad S.
A2 - Obaidat, Mohammad S.
A2 - Davoli, Franco
A2 - Hsiao, Kuei-Fang
A2 - Nicopolitidis, Petros
A2 - Guo, Yu
PB - Institute of Electrical and Electronics Engineers Inc.
Y2 - 10 July 2023 through 12 July 2023
ER -