Quantitative Prognosen von Zeitreihen in modernen Software Tools

Abstract

Quantitative forecasts for time series can lead to optimized results when applied correctly. Unfortunately, controllers often lack the necessary methodological knowledge, and the market for corresponding programs with forecasting functions is opaque. Therefore, this work generally deals with quantitative forecasts for time series and their implementation in leading software tools. However, the introduction of new applications requires resources, and their integration into existing structures can be difficult. For this reason, approaches to implementing forecasting methods in Power BI are also explored, as Power BI is already widely used and often well integrated into existing structures. The work is divided into four parts. In the first and second parts, forecasting itself is explained with the help of scientific literature, followed by a detailed explanation of quantitative methods. In the third part, application books and writings from providers are evaluated to find out which methods are used in the programs. Finally, the last part is devoted to the implementation of forecasting models in Power BI. During the work, it quickly became apparent that there is a lot of different quantitative forecasting methods and several variations within these models. Unfortunately, there is no clear distinction and designation in the literature, which made the evaluation difficult. It was also found that there are only a few empirical studies that deal with the distribution or application of the various methods in companies. In the third part of the work, the analysis of leading software solutions revealed that there are many different applications with forecasting functions. After selecting four programs, it was found that the two planning programs, SAC and Oracle EPM, mainly use statistical methods, while the advanced analytics solutions, Amazon Forecast and R, include newer algorithms. Finally, in the last part, four approaches were identified and directly applied in Power BI to implement forecasting methods in the reporting tool. Although the objectives could largely be achieved, the question remains at the end whether the designed approaches are practical and whether the conveyed methodological knowledge falls within the competency area of controllers. The literature discusses whether activities such as data preparation and the modelling of statistical models fall within the controller's tasks or are already being taken over by other professional fields.

Date of Award	2024
Original language	German (Austria)
Supervisor	Peter Hofer (Supervisor)