Towards KDE mining approach for multi-agent negotiation

The importance of automated negotiation has increased in recent years, because the growing interest in many fields. The aim of such interactions is to reach agreements through exchanging the offers and counteroffers. From these historical offers, mining methods pays special attention for two issues, first because its ability discover and extract knowledge about the opponent preferences, second, its ability to repeat the above process as many times as deemed necessary. As a result it can increase the sensitivity for the opponent, arriving to the agreement in minimum time due to fewer rounds in competitive, win-win relationship. The proposed algorithm uses a non-parametric method, kernel density estimator (KDE), with trade-off to generate offers which are suitable for both agents. Experimental studies have been performed for comparing this algorithm with human and many other algorithms.