Interpreting Attention Models: LSTM vs. CNN : A case study on customer activation
Koen Weterings, Shir-Lee Kimelman, Stefano Bromuri, Marko van Eekelen · 2020
The service sector seeks effective ways to improve customer interaction by means of data driven approaches. This challenge implies personalizing the types and combinations of interactions to engage and activate a customer or a specific group of customers. Prior research on predicting customer activation proposed to use recurrent neural networks powered with attention models. This contribution expands upon this prior research by suggesting a convolutional model with attention and exploring its explainability concerning customer activation in comparison to recurrent models. The results show that recurrent and convolutional models perform similarly in terms of precision and recall. However, the attention models of the recurrent and convolutional architectures focus on different aspects of the customer data, providing different perspectives on customer-based interaction.