دانلود رایگان مقاله پیش بینی حجم تماس های دریافتی در مراکز تماس با شبکه های عصبی مجدد

abstract

Researchers apply Neural Networks widely in model prediction and data mining because of their remarkable approximation ability. This study uses a prediction model based on the Elman and NARX Neural Network and a back-propagation algorithm for forecasting call volumes in call centers. The results can help determine the optimal number of agents necessary to reduce waiting time for customers, enabling profit maximization and reduction of unnecessary costs. This study also compares the performance of the Elman-NARX Neural Network model with the time-lagged feed-forward Neural Network in addressing the same problem. The experimental results indicate that the proposed method is efficient in forecasting the call volumes of call centers.

4. Results and discussion

Table 1 presents the results from the proposed model. The method was applied to several topologies of the Artificial Neural Network the one with the best performance was selected. Table 1 presents the average after obtaining 10 results on each topology. Note that the topologies 4-5-6-1 with 500 epochs show better performance (Minimum MSE) in comparison to other topologies. As there were no capable models for forecasting incoming call volumes, this study compares the proposed model with a base model such as the time-lagged feed-forward network, which is a base model for short-term traffic forecasting using Neural Networks (Haykin, 1998; Mozer, 1993).