A self-learning predictive algorithm of hostile attack based on the weighted k-means

In order to improve the confrontation level in robot soccer competitions, a predictive algorithm of hostile attack based on the weighted k-means is presented. This algorithm clustered the hostile members with k-means algorithm and weighting calculated the opponent attack center by analyzing hostile members in offensive cluster, then preliminarily predicted hostile attack area; Introducing an adaptive self-learning mechanism to the preliminary predictive result, this algorithm generated the final predictive result by analyzing and optimizing information recurrently in knowledge base. In most cases, it is hard to make timely and accurate predictions about hostile attack during high-speed matches. The algorithm is employed to solve the problem that the defense of robot soccers is deficient in purpose and pertinence. Experiments and competitions proved that this method can raise the predictive accuracy effectively and enhance host defensive effect significantly.