Analíticas de aprendizaje y análisis estadístico implicativo: comparación de técnicas clusters para variables modales

Abstract

In an increasingly digitalized world, data analysis has been gaining relevance day by day due to its study has become a fundamental and essential aspect when providing strategic solutions for decision making, therefore, not using the appropriate techniques for the treatment of this information creates repetitive and, in some cases, even unsolvable processes, causing slowness in the calculations and not taking advantage of resources such as execution time and memory space. Given this problem, this research proposed to perform a comparative analysis of the execution time and memory space used between clustering methods similar to the Implicative Statistical Analysis (ASI) and Learning Analytics (LA). For this purpose, a pre-experimental design with modal variables was used, considering for ASI the callHierarchyTree and callSimilarityTree methods and for LA the hclust. vector, hclustvar and diana methods. The study group was 100000 databases, the type of sampling was simple random with a sample size of 382 categorical databases (up to 10 categories) which were randomly generated. It was demonstrated that there is no significant difference in memory space between the five methods, with respect to execution time it was determined that the methods that take the least time are callHierarchyTree and callSimilarityTree followed by hclustvar and finally the methods that took the most time were hclust, vector and diana. These results presented will serve to use a more efficient clustering method in execution time and memory space when analyzing data, thus allowing to opt for optimal algorithms from the point of view of algorithmic complexity.