A carregar...
Publicação
Data-driven marketing in the banking sector : the role of machine learning and interactive interfaces in decision making
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 2.26 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
À medida que os avanços tecnológicos transformam as indústrias, o setor bancário também precisa de se adaptar. Apesar da crescente disponibilidade de dados, os bancos comerciais continuam a encontrar dificuldades na integração destes nas suas estratégias de tomada de decisão, especialmente devido às preocupações com a segurança operacional e rentabilidade. Este estudo aborda a seguinte questão de investigação: Como podem os bancos comerciais integrar eficazmente a análise de dados nos seus processos de decisão e melhorar a eficiência das suas estratégias baseadas em dados? Para responder a esta questão, foi analisado um conjunto de dados público e desenvolvido um modelo preditivo. Após uma análise comparativa entre os algoritmos XGBoost e Regressão Logística, XGBoost foi escolhido devido ao seu melhor desempenho. Apesar de apresentar uma accuracy 7.15% inferior comparativamente a outras versões XGBoost testadas, o modelo final foi escolhido devido à sua capacidade de identificar de forma mais eficaz a classe positiva. Algo considerado essencial em campanhas de marketing onde a identificação de potenciais clientes é o fator mais relevante. Um dos principais desafios dos modelos de Machine Learning (ML) é a falta de transparência e aplicabilidade. Esta investigação abordou estas limitações, através da criação de uma interface adaptativa que processa dados em tempo real e inclui um simulador de previsões, capaz de dar resposta às necessidades de diferentes stakeholders. A sinergia entre o modelo preditivo e a interface proporciona aos utilizadores insights claros sobre o funcionamento do modelo, resolvendo, assim, a questão da "black box” frequentemente associada a ML.
As technological advancements reshape industries, the banking sector must also evolve. Despite the increasing availability of data, commercial banks often struggle to integrate it into their decision-making frameworks, mainly due to concerns about operational safety and profitability. A key research question is addressed during this study: How can commercial banks effectively integrate data analytics into their decision-making frameworks and further enhance the efficiency of data-driven strategies? To address this, a public dataset was analyzed, and a predictive model was developed. After an initial comparison between XGBoost and Logistic Regression algorithms, XGBoost was selected due to its superior performance. Despite having a 7.15% lower accuracy against the other tested XGBoost versions, the chosen model was still preferred due to its better positive class predictions. Positive class predictions are a crucial factor in marketing campaigns where identifying potential customers is more important than overall accuracy. Machine Learning (ML) models are often criticized due to their lack of interpretability and usability. As such, designing an interface is crucial to meet the needs of various stakeholders, including marketing professionals and call center agents. This research addressed these challenges by developing an adaptive interface capable of real-time data analysis, and the implementation of a real-time prediction’s simulator. The synergy between the predictive model and the interface, provides users with meaningful insights into the model’s reasoning, addressing the 'black box' concerns often associated with ML.
As technological advancements reshape industries, the banking sector must also evolve. Despite the increasing availability of data, commercial banks often struggle to integrate it into their decision-making frameworks, mainly due to concerns about operational safety and profitability. A key research question is addressed during this study: How can commercial banks effectively integrate data analytics into their decision-making frameworks and further enhance the efficiency of data-driven strategies? To address this, a public dataset was analyzed, and a predictive model was developed. After an initial comparison between XGBoost and Logistic Regression algorithms, XGBoost was selected due to its superior performance. Despite having a 7.15% lower accuracy against the other tested XGBoost versions, the chosen model was still preferred due to its better positive class predictions. Positive class predictions are a crucial factor in marketing campaigns where identifying potential customers is more important than overall accuracy. Machine Learning (ML) models are often criticized due to their lack of interpretability and usability. As such, designing an interface is crucial to meet the needs of various stakeholders, including marketing professionals and call center agents. This research addressed these challenges by developing an adaptive interface capable of real-time data analysis, and the implementation of a real-time prediction’s simulator. The synergy between the predictive model and the interface, provides users with meaningful insights into the model’s reasoning, addressing the 'black box' concerns often associated with ML.
Descrição
Palavras-chave
Banking sector Customer conversion Data-driven marketing Machine learning Predictive models Streamlit Conversão de clientes Marketing baseado em dados Modelos preditivos Setor bancário
Contexto Educativo
Citação
Editora
Licença CC
Sem licença CC