Forecasting Call Center Arrivals Using XGBoost Combined with Consecutive and Periodic Lookback

Tartuk M., Nurdağ T. F., Acar V., Erdem S., Akay M. F., Abut F.

Eastern Anatolian Journal of Science, cilt.8, sa.1, ss.20-25, 2022 (Hakemli Dergi)

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 8 Sayı: 1
  • Basım Tarihi: 2022
  • Dergi Adı: Eastern Anatolian Journal of Science
  • Sayfa Sayıları: ss.20-25
  • Çukurova Üniversitesi Adresli: Evet

Özet

For companies operating in the call center service sector, it is essential to plan and manage call center employees regularly and optimize the costs. Therefore, agent planning needs to be performed in an optimum way in the call center sector. To make customer representative planning, information on the number of incoming calls is needed to forecast call counts. This study aims to forecast the number of calls using the Extreme Gradian Boosting (XGBoost) combined with consecutive and periodic lookback to be able to plan the number of representatives at specified intervals per operation in the call center sector. Models based on Moving Average (MA) have also been developed for comparison purposes. Mean Absolute Error (MAE) has been used to evaluate the performance of forecast models whereas the generalization errors of the models were evaluated using 80/20 split for training and testing. Forecasts were generated in daily format for four different weeks. The results show that XGBoost performs better than MA for all four different weeks and produces predictions within limits of acceptable accuracy.

For companies operating in the call center service sector, it is essential to plan and manage call center employees regularly and optimize the costs. Therefore, agent planning needs to be performed in an optimum way in the call center sector. To make customer representative planning, information on the number of incoming calls is needed to forecast call counts. This study aims to forecast the number of calls using the Extreme Gradian Boosting (XGBoost) combined with consecutive and periodic lookback to be able to plan the number of representatives at specified intervals per operation in the call center sector. Models based on Moving Average (MA) have also been developed for comparison purposes. Mean Absolute Error (MAE) has been used to evaluate the performance of forecast models whereas the generalization errors of the models were evaluated using 80/20 split for training and testing. Forecasts were generated in daily format for four different weeks. The results show that XGBoost performs better than MA for all four different weeks and produces predictions within limits of acceptable accuracy.