一种具有动态邻域特点的自适应最近邻居算法
发布时间：2018-11-22 06:52
【摘要】：传统的最近邻居算法主要分为k-最近邻居和逆最近邻居,然而二者均在邻域参数选择问题中饱受困扰。在这两种思想的基础上,提出一种具有动态邻域特点的最近邻居算法——自然邻居,并围绕其概念与特性形成了一套有效的方法。该算法从根本上克服了传统最近邻居思想在任意形状(如流型)数据集中参数选择的难题,摆脱了传统方法的参数依赖,并且取得了极佳的效果。自然邻居思想具有完善的理论模型和详细的实现算法,并且经验证其具有很强的鲁棒性和适应性。
[Abstract]:The traditional nearest neighbor algorithm is mainly divided into k- nearest neighbor and inverse nearest neighbor. However, both of them are troubled in the neighborhood parameter selection problem. On the basis of these two ideas, a nearest neighbor algorithm with dynamic neighborhood characteristics, natural neighbor, is proposed, and a set of effective methods are formed around its concept and characteristics. This algorithm fundamentally overcomes the problem of parameter selection in arbitrary shape (such as flow pattern) data set based on traditional nearest neighbor idea, and gets rid of the parameter dependence of traditional method, and achieves excellent results. The idea of natural neighbor has perfect theoretical model and detailed implementation algorithm, and it has strong robustness and adaptability.
【作者单位】： 重庆师范大学计算机与信息科学学院;重庆大学软件理论与技术重庆市重点实验室;
【基金】：重庆市自然科学基金(cstc2013jcyjA40049) 重庆师范大学基金项目(17XLB003)资助
【分类号】：TP301.6
本文编号：2348412
[Abstract]:The traditional nearest neighbor algorithm is mainly divided into k- nearest neighbor and inverse nearest neighbor. However, both of them are troubled in the neighborhood parameter selection problem. On the basis of these two ideas, a nearest neighbor algorithm with dynamic neighborhood characteristics, natural neighbor, is proposed, and a set of effective methods are formed around its concept and characteristics. This algorithm fundamentally overcomes the problem of parameter selection in arbitrary shape (such as flow pattern) data set based on traditional nearest neighbor idea, and gets rid of the parameter dependence of traditional method, and achieves excellent results. The idea of natural neighbor has perfect theoretical model and detailed implementation algorithm, and it has strong robustness and adaptability.
【作者单位】： 重庆师范大学计算机与信息科学学院;重庆大学软件理论与技术重庆市重点实验室;
【基金】：重庆市自然科学基金(cstc2013jcyjA40049) 重庆师范大学基金项目(17XLB003)资助
【分类号】：TP301.6
【相似文献】
相关博士学位论文 前1条
1 吴珊珊;数据流频繁项挖掘及相关性分析算法的研究[D];浙江大学;2017年
相关硕士学位论文 前6条
1 贾yN恺;基于深度特征学习的目标检测与跟踪算法研究[D];西安科技大学;2017年
2 张灿龙;不确定DM-chameleon聚类算法在滑坡危险性预测的研究及应用[D];江西理工大学;2017年
3 刘星晨;基于集成半监督学习框架的电影视频总结算法研究[D];天津大学;2016年
4 韩琼;基于边界度模型的聚类技术研究[D];郑州大学;2017年
5 邵星星;基于尺度不变特征变换的虹膜识别技术研究[D];郑州大学;2017年
6 刘聪;基于k-匿名的轨迹隐私保护技术研究[D];江西理工大学;2017年
，本文编号：2348412
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2348412.html
