Análisis y predicción del comportamiento de la operación de transporte de carga automotor usando técnicas de minería de datos

Abstract

En la actualidad Colombia presenta fuertes falencias en cuanto al manejo de los flujos de información Logística, que se encuentra segmentada, dispersa o ausente. Situación que limita la formulación de políticas públicas y planes de acción específicos enfocados a la optimización del sector logístico desde una visión integral de la cadena de abastecimiento, y como apoyo al incremento de competitividad y productividad. De manera particular no se cuenta con información detallada del funcionamiento de los principales corredores logísticos del país, desconociendo detalles importantes como lo son cantidades y tipos de carga movilizadas, tramos críticos que los componen, frecuencias de viajes dentro de los corredores, etc. Los sistemas de información disponibles presentan fuertes falencias en el apoyo a los procesos de planeación y toma de decisiones, ya que no incorporan herramientas descriptivas ni predictivas de la información que permitan realizar monitoreo y extracción de información relevante de grandes volúmenes de datos. Por lo tanto, herramientas analíticas como la minería de datos se convierten en una excelente alternativa para apoyar los procesos de generación de información para la planeación y la toma de decisiones en proyectos de transporte, impactando de manera directa en la logística y competitividad del país. La presente propuesta presenta el resultado del desarrollo de un sistema prototipo de apoyo a la toma de decisiones basado en técnicas de minería de datos que permite solventar las necesidades de información críticas en torno al comportamiento de los corredores logísticos en Colombia. Para la construcción del prototipo se basó en la metodología CRISP-DM 1.0 y los datos provenientes del sistema de información Registro Nacional de Despacho de Carga (RNDC).

Currently Colombia has strong shortcomings in the management of logistics information flows, which is segmented, dispersed or absent. This situation limits the formulation of public policies and specific action plans aimed at optimizing the logistics industry from a holistic view of the supply chain and to support the increased competitiveness and productivity. In particular there is little detailed information on the functioning of the main logistics corridors in the country, ignoring important details such as quantities and types of cargo mobilized, critical sections that compose frequencies travel inside corridors, etc. Information systems available have strong weaknesses in supporting planning processes and decision making, and which do not incorporate predictive and descriptive tools of information that allow for monitoring and extracting relevant information from large volumes of data. Therefore, analytical tools such as data mining become an excellent alternative to support the process of generating information for planning and decision-making on transport projects directly impacting on logistics and competitiveness. This proposal presents the result of the development of a prototype system to support decision-making based on data mining techniques enabling meet the needs of critical information about the behavior of the logistic corridors in Colombia. For the construction of the prototype based on the methodology CRISP-DM 1.0 and data from the National Register Information System Load Dispatch (RNDC).

Abstract. Currently Colombia has strong shortcomings in the management of logistics information ows, which is segmented, dispersed or absent. This situation limits the formulation of public policies and speci�c action plans aimed at optimizing the logistics industry from a holistic view of the supply chain and to support the increased competitiveness and productivity. In particular there is little detailed information on the functioning of the main logistics corridors in the country, ignoring important details such as quantities and types of cargo mobilized, critical sections that compose frequencies travel inside corridors, etc. Information systems available have strong weaknesses in supporting planning processes and decision making, and which do not incorporate predictive and descriptive tools of information that allow for monitoring and extracting relevant information from large volumes of data. Therefore, analytical tools such as data mining become an excellent alternative to support the process of generating information for planning and decision-making on transport projects directly impacting on logistics and competitiveness. This proposal presents the result of the development of a prototype system to support decision-making based on data mining techniques enabling meet the needs of critical information about the behavior of the logistic corridors in Colombia. For the construction of the prototype based on the methodology CRISP-DM 1.0 and data from the National Register Information System Load Dispatch (RNDC).