Identificación de patrones que influyen en los bajos rendimientos industriales en la fabricación de azúcar
Year: 2024
Type: article
Abstract: The computerization of sugar industry processes generates a large amount of data that is increasingly stored over the years. Currently, the application of the programs of the agroindustrial platform existing in the Higher Management Organization for the Sugar Agroindustry (AZCUBA) has guaranteed the speed and quality of harvest information. The Cuban sugar industry requires the implementation of scientific tools and methods that allow the identification of hidden patterns and behaviors in its historical data. This article presents the use of knowledge extraction techniques from data to identify the causes that are influencing low industrial yields. Among the materials used are the databases of ten years of harvest (2010-2019), which each present more than 4 million transactional records and an average of 578 indicators per year. The methodology selected to establish a framework for the life cycle of the data mining process was CRISP-DM. The tool selected to apply the data mining techniques was the KNIME data analysis platform. A predictive analysis of the data was performed, in which symbolic methods were used. The metrics of seven machine learning algorithms were compared: CONJUNCTIVERULE, DECISIONTABLE, RIDOR, FURIA, PART, JRIP, J48, for the selection of features, and the selection of the algorithm for classification was determined. The attributes that influence low industrial performance are obtained and validated. The bases for a deeper analysis of the necessary organizational and control measures are created, with the objective of losing sugar in the industrial process. It is recommended to perform a prescriptive analysis of the data, to predict logistics scenarios.
Institution: Polytechnic José Antonio Echeverría
Language: Spanish
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7220074091