My SciELO
Services on Demand
Article
Indicators
- Cited by SciELO
Related links
- Similars in SciELO
Share
Revista Universidad y Sociedad
On-line version ISSN 2218-3620
Abstract
RODRIGUEZ LEON, Ciro and GARCIA LORENZO, María Matilde. ADAPTATION TO A METHODOLOGY OF DATA MINING FOR APPLYING TO UN-SUPERVISED PROBLEMS TYPE ATTRIBUTE-VALUE. Universidad y Sociedad [online]. 2016, vol.8, n.4, pp. 43-53. ISSN 2218-3620.
The amount of any kind of stored data is going in an exponential increment. That is why it is needed to create efficient procedures to manipulate this data and extract knowledge from them. Data mining is in charge of this type of process and to make their procedures less complex. Methodologies have been designed to guide them. As these methodologies are general they do not describe important issues as techniques and algorithms to be used in each period. In the present research, after a comparative study, CRISP-DM methodology is selected to be adapted to un-supervised problems type attribute-value. In this way, by reducing the application domain, it is achieved a deeper specification level in each of the six phases which were originally proposed, so time of specialists with the purpose of doing this kind of activity,is saved. To demonstrate the use of this adaptation and its successful results, it is applied to a real case study, consisting in a group of type 2 diabetic patients in which satisfactory results are achieved after an independent analysis by sex. The groups found represent different levels of risk factors in the disease evolution who improve their prevention process and diagnosis.
Keywords : Data mining; CRISP-DM; clustering; validation index; diabetes.