Automation of Rolling Forecasting: A Case Study in the Norwegian Process Industry
Master thesis
Permanent lenke
https://hdl.handle.net/11250/3020333Utgivelsesdato
2022Metadata
Vis full innførselBeskrivelse
Full text not available
Sammendrag
Artificial intelligence have fascinated scientist with the ability of imitating human behaviourand mindset through mathematical computing. It can also be argued that artificial intelligencehave the potential to surpass human intelligence. The opportunities within utilizingthe technology can be combined with endless combinations; however, attributes for an algorithmcomes at a cost that can affect an organization’s business strategy and technicalinfrastructure. It is therefore important that compromises are taken in a well reflectedmanner and appropriate conditions that suits the dynamic market that the organization liewithin.
This master’s thesis aims to illuminate opportunities and limitation associated with the implementationof an artificial intelligence driven rolling forecasts in a Norwegian fiberglassmanufacturing plant using a qualitative research methodology that consists of interviewswith key personnel from the organization. Experimental models have been developed toachieve an understanding of the implementation scope through a comparison of an autoregressiveintegrated moving average and long short-term memory predictive models.
The study results have been compared to a theoretical framework revolving around the mainthemes of the thesis: financial forecasting, management accounting & control and artificialintelligence where the following outcomes have been identified: (1) implementation of thetechnology require prerequisites in the form of integrated business processes, (3) forecastingand budgeting must have distinguished roles, (2) the management have to increase technologicalfocus within business activities, (4) a systematic approach towards data collectionhave to be established, (5) optimal balance between objective and subjective forecastingis preferred and (6) prior manual activities that the predictive model is dependent on canincrease the risk of inaccurate forecasts.