The impact of data quality on the effectiveness of machine learning models in the bakery industry in the context of big data

Abstract

The article examines the impact of data quality on the effectiveness of machine learning models in the bakery industry in the context of big data. The relevance of the topic is due to the growing role of data analytics in optimizing bakery production and the need to ensure the reliability of the predictive models used. The aim of the work is to identify key data quality parameters that determine the accuracy and practical applicability of machine learning models in the baking industry. A set of methods was used in the study, including statistical analysis of arrays of bakery production data, expert interviews (n=20) and comparative testing of models on training samples of different quality. It is established that: 1) completeness, accuracy and consistency of data are key factors affecting the generalizing ability of models; 2) the use of data preprocessing (purification, transformation) allows to increase the accuracy of predictions of the output of bakery products by an average of 10-15%; 3) models trained on high-quality data demonstrate three times higher stability in the test sample; 4) the quality of forecasting key indicators of the baking process in adaptive models can exceed existing standards by 8-12%. The results confirm the critical importance of data quality management for realizing the potential of machine learning in the bakery industry. A methodology for auditing the quality of technological data of bakeries is proposed, focused on the specifics of modeling and optimization tasks. Further research is related to the development of infrastructure and management solutions to ensure data quality in the context of digitalization of bakery production.