Institutional Repository of Vilnius University
Įeinančių skambučių srauto prognozavimas /
Abstract
dc:descriptionThe main goal of business contact center is to ensure good customer service and experience using as little employees and other resources as possible. One way to achieve it - is to forecast incoming customer calls for different time periods. Then using the predicted data calculate how many employees are needed when forming weekly and monthly work schedules. The purpose of this paper is to forecast incoming ̨ calls ̨for each working hour and for each working day while using different methods. The main goals of this work are forecasting incoming calls using the time series (ARIMA) model, machine learning (K nearest neighbors, random forests and support vectors) regression models, and applying the recurrent neural network (LSTM) model, evaluating and comparing them. In the research, all used models are applied for two differently aggregated datasets. The first dataset is the number of incoming calls for each working hour, and the second dataset is the total number of incoming calls for each day. The conducted study showed that the daily number of calls is most accurately predicted when RNN methods are used. The KNN and RF models perform similar predictions for both datasets, but the RF model showed more accurate predictions according to the MAPE criteria. The SVR machine learning model prediction was the worst of all the models that were used. By comparing models, we can say that the LSTM model of recurrent neural networks shows the most accurate results and is the most promising in this field given its adaptability to process and analyze sequential data over several time steps.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Eigelytė, Ieva,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD210644688&prefLang=en_US