[4Xin1] Poster session 2

[4Xin1-10] Predicting Order Timing Utilizing Temporal Expressions in Sales Activity Reports

〇Masaya Tsunokake¹, Yuta Koreeda¹, Takeshi Homma¹, Teppei Inoue¹, Yuhei Niwa¹ (1.Hitachi, Ltd.)

Keywords:Time-to-Event Prediction, Natural Language Processing, Sales Force Automation/Marketing Automation, Temporal Expression Extraction

If the order timing in sales activities can be predicted with a high degree of accuracy, the activities can be managed more optimally. In addition, the processes of procurement, manufacturing, and development for product and system development required after contracting can also be preemptively managed. This leads to improvements in management efficiency and profitability. One possible method for predicting the order timing is to predict the order date based on metadata that indicates the characteristics of the project. However, this method cannot consider the progress status and schedule of actual sales activities, which change over time, which may results not high accurate prediction. This study aims to predict the order timing considering the status of work execution and the schedule of work execution, using sales reports that describe the sales activities. In this paper, we propose the method that uses time expressions in sales reports as clues. The our method identifies when the work associated with the process until the contracting is conducted based on the time expressions, and creates time information and progress status of each process as features for prediction. On the experiment using actual sales activity data, the our method achieved the best prediction performance compared to prediction methods that do not utilize time expressions in sales reports.