Research On Genetic Algorithm Based On Relative Issues Concurrent Negotiation

There are researches from the most simple bilateral negotiation on single issue to complex negotiation among multilateral on multi issues in the field of intelligent negotiation of e-commerce. Implementation mechanisms of negotiation models are different from each other, covering many technologies and methods, such as grid technology, concurrent method, fuzzy logic control method, case-based reasoning technology, game theory method and genetic algorithm.The existing research has some achievements for multilateral multi-issue negotiations, but there are also some flaws. Most of the studies didn't inspect the association among issues, and they supposed issues' weights were given fixed, and issues'initial weights were specified one-time in a treaty. Issues proposed by negotiating parties usually have certain correlation among them and some fluctuations of issue weights occur possibly during the treaty progress when compared with the treaty form in reality. Therefore exsiting negotiation models'scope of application is narrowed by these constraints.A new multilateral multi issues negotiation algorithm named Relative Issues Genetic Algotithm is proposed to solving these problems. It doesn't only inspect relativity among issues, but also covers dynamic change of weights. There appears the "in-relative and out-independent" pattern by grouping method. Foundation of group weight alteration is gained from the most similar case searched in trade history by case-based reasoning.The concurrent negotiation model is constructed on the foundation of algorithm design. Case-Based Reasoning technology and moving average method for analysing history data are applied in the model to find history recommended value of user's weight. Then the weight decision-making function choses a value as final available weight between user's current data and history value. The measuring rule from issue weight to group weight is given. Meanwhile the method on calculating group utility from issue utility is designed.The model is validated reasonable and feasible through numerical experiment on algorithm and emulation model system analysis. It doesn't only simplify management on secondary issues for users, facilitate users to use model better, but also supports more treaty information besides the best answer on users' demand to help making decisions.