Methodology for Railway Demand Forecasting Using Data Mining
Giovanni Melo Carvalho Viglioni · 2007
After the organization to resolve its operational problems, comes the necessity for systems to support decision making. The research area data mining grows quickly to take care of these new necessities. However, the use of data mining techniques becomes difficult due to the lack of a complete and systematic methodology for the knowledge discovery in database. This dissertation presents a model of the formal process of development of systems of discovery of knowledge in database for the prediction of railroad demand, that includes a systematic and rigorous methodology, which integrates the methodologies: CRISP-DM, SEMMA, FAYYAD, and an interactive environment for the implementation of these systems. The methodology proposal integrates the cited methodologies and was applied in a customer transport request database of MRS Logistica, during the period of Dec, 1 st of 2003 until Oct, 31 st of 2006. This application is main objective was to validate the methodology proposal according to the criteria of the respective company. The conclusions of the case studies allowed us to show the relevance of the MPDF-DM methodology in the forecast of railroad demand.