Time Series Forecasting in the B2B Market

With the advancements in information technology and data availability, forecasting is growingly considered as a critical capability within organisations. Endress+Hauser Flowtec AG (E+H) aims to explore quantitative forecasting models by overcoming cognitive limitations posed by human judgement.

Briatico, Antonio, 2022

Art der Arbeit Bachelor Thesis
Auftraggebende Endress+Hauser Flowtec AG
Betreuende Dozierende Jüngling, Stephan
Keywords Forecasting, Time Series Analysis, Data Exploration, Prophet
Views: 18
Dynamics of business activities often force managers to make decisions based on subjective mental models, reflecting their experiences. To support the decision-making process, it is necessary to investigate quantitative forecasting models that generate interpretable predictions while enabling intuitive parameter adjustments. Also, global economic indicators are included in the analysis to gain a better understanding of external influences.
To introduce quantitative models for generating predictions, the forecasting process is examined based on conceptual foundations of time series analysis. Relevant Machine Learning (ML) and statistical forecasting models are critically analysed and economic indicators as explanatory variables are evaluated. The forecasting model is proposed based on model accuracy and interpretability.
The critical analysis of forecasting models emphasises a regression model provided by “Prophet” that handles non-linear data and time series components automatically. It enables parameter adjustment based on domain knowledge and therefore improves interpretability. Also, the output of the model demonstrates an increase in forecast accuracy compared to commonly used benchmark models by including the Gross Domestic Product (GDP) as an additional regressor. In addition, the empirical research serves as a guidance to standardise the overall approach on generating predictions. The proposed model allows to incorporate data-driven forecasts in the decision-making process and thus counters biased and subjective inputs posed by the AS-IS situation at E+H.
Studiengang: Business Information Technology (Bachelor)
Vertraulichkeit: vertraulich
Art der Arbeit
Bachelor Thesis
Auftraggebende
Endress+Hauser Flowtec AG, Reinach BL
Autorinnen und Autoren
Briatico, Antonio
Betreuende Dozierende
Jüngling, Stephan
Publikationsjahr
2022
Sprache der Arbeit
Englisch
Vertraulichkeit
vertraulich
Studiengang
Business Information Technology (Bachelor)
Standort Studiengang
Basel
Keywords
Forecasting, Time Series Analysis, Data Exploration, Prophet