[1]韩迪,等.增量学习的优化算法在app使用预测中的应用[J].深圳大学学报理工版,2019,36(1):43-51.[doi:10.3724/SP.J.1249.2019.01043]
 HAN Di,LI Wenting,et al.The application of optimization algorithm based on incremental learning in app usage prediction[J].Journal of Shenzhen University Science and Engineering,2019,36(1):43-51.[doi:10.3724/SP.J.1249.2019.01043]
点击复制

增量学习的优化算法在app使用预测中的应用()
分享到:

《深圳大学学报理工版》[ISSN:1000-2618/CN:44-1401/N]

卷:
第36卷
期数:
2019年第1期
页码:
43-51
栏目:
电子与信息科学
出版日期:
2019-01-20

文章信息/Info

Title:
The application of optimization algorithm based on incremental learning in app usage prediction
作者:
韩迪1 2李雯婷3 王庆娟1周天剑1路良刚1
1)北京理工大学珠海学院计算机学院,广东珠海 519000;2)澳门科技大学资讯科技学院,澳门 999078;3)贵州商学院计算机与信息工程学院,贵州贵阳 550014
Author(s):
HAN Di1 2 LI Wenting3 WANG Qingjuan1 ZHOU Tianjian1 and LU Lianggang1
1) Beijing Institute of Technology, Zhuhai, Zhuhai 519000, P.R.China 2) Macau University of Science and Technology, Macao, 999078, P.R.China 3) Guizhou University of Commerce, Guiyang, 550014, Guizhou Province, P.R.China
关键词:
模式识别App使用预测聚类增量学习大数据
Keywords:
pattern recognition app usage prediction clustering incremental learning big data
分类号:
TP 311
DOI:
10.3724/SP.J.1249.2019.01043
摘要:
随着智能手机上app数量的不断增加,准确查询目标app渐趋困难,在手机系统预测下一个将要启动的app日趋必要和重要.利用历史用户数据来预测下一个使用的app算法尚存在两类问题:一是部分算法因未考虑训练数据随着时间增加而增加,导致预测结果的准确度随着时间增加而降低;二是虽然考虑到了增量数据,但增加了因增量数据而重新建模的时间,致总体耗时增加.为减少建模时间,本研究提出Predictor预测系统,利用优化后的增量IkNN模型为用户提供app使用的预测功能.同时本研究通过学习app的特征的上下文关系,设计了聚类有效值(cluster effective value, CEV)策略,采用多维度特征方法来提高分类的准确度,从而提高预测准度.实验结果表明,带有CEV策略的IkNN模型比默认的IkNN模型拥有更稳定的预测准度,其应用模型Predictor能减少建模的时间,同时提高预测准度.
Abstract:
Nowadays it is becoming harder and harder to find an app on a smartphone due to the increasing number of apps installed. So predicting the next app usage quickly and accurately is very important and necessary. There are two kinds of problems in predicting the next app usage according to the usage history of apps. One is that some algorithms do not consider the increment of the training data over time, which reduces the prediction accuracy over time. The other one is that some algorithms consider the increment of the training data over time, but they rebuild the model using all historical data once the increased data reaches to a limit, thus greatly increases the remodeling time. To reduce the remodeling time, we utilize an incremental k-nearest neighbors (IkNN) model algorithm to implement a solution, called Predictor. When the IkNN model is used for predicting the next app usage, a new problem is found. Modeling with training data, the classification accuracy is reduced with the increase of the number of features of an app. After studying relationship between the context features of an app, we design a cluster effective value (CEV), which can compensate the error induced by multidimensional features, to improve the classification accuracy, thus to improve the prediction accuracy. It is shown that the IkNN Model algorithm with CEV has a higher and more stable prediction accuracy than that of the algorithm without CEV. Large-scale experiments show that the Predictor can reduce remodeling time and improve the prediction accuracy.

相似文献/References:

[1]胡涛,郭宝平,郭轩.基于游程分析轮廓提取算法的改进[J].深圳大学学报理工版,2009,26(4):405.
 HU Tao,GUO Bao-ping,and GUO Xuan.An improved run-based boundary extraction algorithm[J].Journal of Shenzhen University Science and Engineering,2009,26(1):405.
[2]朱安民,陈燕明.基于特征点一致性约束的实时目标跟踪算法[J].深圳大学学报理工版,2013,30(No.3(221-330)):228.[doi:10.3724/SP.J.1249.2013.03228]
 Zhu Anmin and Chen Yanming.A real-time target tracking algorithm based on the consistency constraint of feature points[J].Journal of Shenzhen University Science and Engineering,2013,30(1):228.[doi:10.3724/SP.J.1249.2013.03228]
[3]柳伟,陈旭,梁永生,等.基于时空显著图的可伸缩视频码率控制方法[J].深圳大学学报理工版,2013,30(No.5(441-550)):462.[doi:10.3724/SP.J.1249.2013.05462]
 Liu Wei,Chen Xu,Liang Yongsheng,et al.Rate control for scalable video coding based on spatiotemporal saliency map[J].Journal of Shenzhen University Science and Engineering,2013,30(1):462.[doi:10.3724/SP.J.1249.2013.05462]
[4]罗雪晖,李霞,张基宏.支持向量机及其应用研究[J].深圳大学学报理工版,2003,20(3):40.
 LUO Xue-hui,LI Xia and ZHANG Ji-hong.Introduction to Support Vector Machine and Its Applications[J].Journal of Shenzhen University Science and Engineering,2003,20(1):40.
[5]刘翠响,袁香伟,王宝珠,等.最小均衡化后的行人重识别[J].深圳大学学报理工版,2019,36(4):447.[doi:10.3724/SP.J.1249.2019.04447]
 LIU Cuixiang,YUAN Xiangwei,WANG Baozhu,et al.Minimum equalization for pedestrain re-identification[J].Journal of Shenzhen University Science and Engineering,2019,36(1):447.[doi:10.3724/SP.J.1249.2019.04447]
[6]刘润奇,贺兴时,南夷非,等.网络多媒体数据中舆情关联主题的挖掘方法[J].深圳大学学报理工版,2020,37(1):72.[doi:10.3724/SP.J.1249.2020.01072]
 LIU Runqi,HE Xingshi,NAN Yifei,et al.Mining method of public opinion related topic in network multimedia data[J].Journal of Shenzhen University Science and Engineering,2020,37(1):72.[doi:10.3724/SP.J.1249.2020.01072]
[7]黄林,常健,杨帆,等.基于改进k-means的电力信息系统异常检测方法[J].深圳大学学报理工版,2020,37(2):214.[doi:10.3724/SP.J.1249.2020.02214]
 HUANG Lin,CHANG Jian,YANG Fan,et al.An anomaly detection method for electric power information system based on improved k-means[J].Journal of Shenzhen University Science and Engineering,2020,37(1):214.[doi:10.3724/SP.J.1249.2020.02214]

更新日期/Last Update: 2019-01-30