An Artificial Neural Network Model for Project Effort Estimation

Şengüneş B., Öztürk N.

SYSTEMS, vol.11, no.2, 2023 (SSCI) identifier identifier

  • Publication Type: Article / Article
  • Volume: 11 Issue: 2
  • Publication Date: 2023
  • Doi Number: 10.3390/systems11020091
  • Journal Name: SYSTEMS
  • Journal Indexes: Social Sciences Citation Index (SSCI)
  • Keywords: artificial neural network, project effort estimation, customized machine development, SOFTWARE-DEVELOPMENT, REGRESSION
  • Bursa Uludag University Affiliated: Yes


Estimating the project effort remains a challenge for project managers and effort estimators. In the early phases of a project, having a high level of uncertainty and lack of experience cause poor estimation of the required work. Especially for projects that produce a highly customized unique product for each customer, it is challenging to make estimations. Project effort estimation has been studied mainly for software projects in the literature. Currently, there has been no study on estimating effort in customized machine development projects to the best of our knowledge. This study aims to fill this gap in the literature regarding project effort estimation for customized machine development projects. Additionally, this study focused on a single phase of a project, the automation phase, in which the machine is automated according to customer-specific requirements. Therefore, the effort estimation of this phase is crucial. In some cases, this is the first time that the company has experienced the requirements specific to the customer. For this purpose, this study proposed a model to estimate how much work is required to automate a machine. Insufficient effort estimation is one of the main reasons behind project failures, and nowadays, researchers prefer more objective approaches such as machine learning over expert-based ones. This study also proposed an artificial neural network (ANN) model for this purpose. Data from past projects were used to train the proposed ANN model. The proposed model was tested on 11 real-life projects and showed promising results with acceptable prediction accuracy. Additionally, a desktop application was developed to make this system easier to use for project managers.