动态环境下基于微粒群优化算法的数据分类方法研究

摘要

The data flow in dynamic environment prone to concept drift phenomenon, gradually along with the data, implicit knowledge in a certain extent can appear in the data change, the current data classification method to dynamic update, not suitable for the classification of the data in dynamic environment.For this, a new data classification method was put forward based on particle swarm optimization algorithm, through the method of K-means classifying data in dynamic environment.The particle swarm optimization algorithm was introduced, all individuals will be considered add no volume of particles in search space.In the search space flight at a certain speed, the speed can be through its own dynamic adjustment and adjacent particles flying experience, through some rules to update the local optimal value of new particles, using the optimized particle swarm algorithm for data classification.The experimental results show that the proposed method classification performance, high accuracy in real time.%动态环境下数据流容易出现概念漂移现象.随着数据的逐渐到达,隐含在数据中的知识在一定程度上会出现改变,当前数据分类方法无法进行动态更新,不适于动态环境下数据的分类.为此,提出一种新的基于微粒群优化算法的数据分类方法,通过K-means方法对动态环境下的数据进行分类.介绍了微粒群优化算法,将所有个体看作d维搜索空间中没有体积的微粒,在搜索空间中以某一速度飞行,该速度可通过其自身及相邻微粒的飞行经验进行动态调整.通过某种规则对新微粒的局部最优值进行更新,利用优化后的微粒群算法实现数据分类.实验结果表明,所提方法分类性能优,实时准确率高.