TBIT: Transformer-based Business Prediction Model Integrating Work Order Textual Contextual Information
Conference
Yin, M, Zhang, W, Guo, S et al. (2021). TBIT: Transformer-based Business Prediction Model Integrating Work Order Textual Contextual Information
. 103-108. 10.1145/3490322.3490339
Yin, M, Zhang, W, Guo, S et al. (2021). TBIT: Transformer-based Business Prediction Model Integrating Work Order Textual Contextual Information
. 103-108. 10.1145/3490322.3490339
Accurate business prediction provides a good understanding of individual business trends and has a significant impact on reducing overall operating costs and ensuring a high level of customer service quality. Currently, artificial intelligence and big data technologies have been widely used in various businesses of electric power customer service, but how to effectively use the heterogeneous information in customer service data for business prediction has become a major challenge. To solve this problem, this paper proposes a Transformer-based Business Prediction Model Integrating Work Order Textual Contextual Information (TBIT) to perform customer service business prediction. The model first uses one-dimensional convolution and self-attention mechanism to mine the semantic contexts contained in the work order text information. Then, the Transformer is used to fuse the customer's dynamic preference information with the work order text context, and to mine the global time sequence dependence of work order data, so as to realize customer service business prediction. The final experiments show that the TBIT model can perform representation learning on work order data more effectively compared to the baseline models, and also show superior performance in the customer service business prediction task.
