Car sales analysis in the Nordic Countries
Martinez, Luis (2023)
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2023052212601
https://urn.fi/URN:NBN:fi:amk-2023052212601
Tiivistelmä
Sales forecasting is an essential component of business intelligence and, artificial intelligence and predictive analytics are now essential tools for companies to predict market trends and forecast sales volumes.
In the automotive industry, where production processes are extremely complex and the logistic operations until the products are retailed and handed over to end customers are very long, the ability to predict future demand and accommodate such prediction to production planning and supply chain processes is even more crucial than in any other industry.
In this work, some classical time-series forecasting techniques are compared with more advanced machine learning methods analyzing their performance in predicting the number of sales of a carmaker in the Nordic countries and studying the correlation between socio-economic indicators and the volume of sales for this specific brand.
Multiple experiments are conducted and evaluated following a quantitative approach and the effectiveness of the models is scrutinized using different performance metrics leading to the conclusion that the introduction of exogenous variables as inputs for ML algorithms can improve the forecasting results overcoming traditional models in the 4 Nordic countries analyzed.
In the automotive industry, where production processes are extremely complex and the logistic operations until the products are retailed and handed over to end customers are very long, the ability to predict future demand and accommodate such prediction to production planning and supply chain processes is even more crucial than in any other industry.
In this work, some classical time-series forecasting techniques are compared with more advanced machine learning methods analyzing their performance in predicting the number of sales of a carmaker in the Nordic countries and studying the correlation between socio-economic indicators and the volume of sales for this specific brand.
Multiple experiments are conducted and evaluated following a quantitative approach and the effectiveness of the models is scrutinized using different performance metrics leading to the conclusion that the introduction of exogenous variables as inputs for ML algorithms can improve the forecasting results overcoming traditional models in the 4 Nordic countries analyzed.