SPDW-Miner: um método para a execução de processos de descoberta de conhecimento em bases de dados de projetos de software

• Autor(a) principal: Figueira, Fernanda Vieira
• Data de Publicação: 2008
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional PUCRS
• Texto Completo: http://hdl.handle.net/10923/1503
• topic: INFORMÁTICA; MINERAÇÃO DE DADOS (INFORMÁTICA); DATA WAREHOUSE

Software organizations aim at improving their Software Development Process (SDP) targeting the quality assessment of their processes and products. They adopt software maturity models to achieve this. Maturity models define quality measuring should be done through a metrics program. The defined metrics must be collected and stored properly, maintaining the history of the organizational quality data. However, measuring alone is not enough. Stored data must be useful to support SDP quality maintenance. To do that, maturity models suggest the use of statistical and analytical techniques. The goal is to make feasible the quantitative understanding of the metrics. Data mining techniques are useful in this scenario as an approach able to improve analytical and predictive capabilities on estimations and performance of SDP. This work introduces a method of performing KDD process, named SPDW-Miner, oriented to software metrics prediction. It is proposed a KDD process that incorporates the SPDW+ data-warehousing environment. Such method is composed by a set of steps that guide users to apply the whole KDD process. In special, instead of considering DW as an intermediate step, SPDW-Miner adopts it as a reference to rule its execution. It is specified all KDD process steps: defining the mining goal; extracting a preparing data; data mining and results optimization. The contribution of this work is the establishing of a KDD process, in a proper, user-comfortable detail level. It enables organizational users can to adopt it as a reference guide to knowledge discovery.