Estudio comparativo de algoritmos de predicción para la minería de datos aplicado al area académica FIE-ESPOCH.

Abstract

Comparative Study of data mining algorithms to determine the best algorithm performance applied to the academic area of the Faculty of computer science and electronics of Escuela Superior Politecnica de Chimborazo. To validate the hypothesis on the performance of the chosen algorithms, the scientific method and the following elements such as: a portable computer and software tools were used as well as Microsoft SQL Server, Microsoft Data Quality, Microsoft Data Tools, Microsoft Analysis Services, and the observation technique to carry out the measurements. The z-test for two-tailed, and a significance level of 5% were applied on the use of a Central Processing Unit (CPU), employ of the Random Access Memory (RAM), response time and accuracy of the algorithms. The result of the analysis showed that the decision tree algorithm has better performance than the logistic regression algorithm with 98.92% on 95.70% of performance. Under the guidance of the methodology Cross Industry Standard Process for Data Mining CRISPDM the project of data mining was implemented on the information of academic indicators of the Faculty of computer science and electronics, making use of the decision tree algorithm of Microsoft, to get the patterns of behavior required. It is concluded that the decision tree algorithm had better performance than the logistic regression algorithm on the academic data of the Faculty of computer science and electronics. An initial study of information prior to the choice of algorithms is recommended to avoid the discarding of data types in a data mining project.