Multiagent supported product management in e-commerce

The most important problem of B2C e-commerce websites is the product management. The variety of product types increase the complexity of the decisions in this area. Temporary solutions that are used put the future of the e-business sites at risk. For this reason, the importance of having a holistic, fast and active solution is increased for these types of businesses. The solution for this kind of problems need the smart, autonomous, electronic information communication infrastructure. The most important representative for these systems nowadays is the multi-agent systems (MAS). In this study, a multi agent system that is supported by genetic algorithm with dynamic decision infrastructure for B2C e-commerce sites, is developed. In order to model the proposed MAS, the Promethous methodology was used. the developed model is tested with the data from ABC, an e-commerce website. in Turkey. In order to support MAS decisions of product management like product selection for favorite products, definining of purchasing policies SOM (self organizing maps), basic regression, artificial neural network and k-means cluster analysis methods. As a result of this Phd study some proposals were given to B2C websites and a MAS model which uses the cluster based demand forecasting (CBDF) modelling. Proposed cluster based and MAS supported product management highly improved performance of demand forecasting and peformances of other decides.