信息检索中基于智能优化算法的数据融合方法的研究

发布时间：2018-01-19 07:06

本文关键词： 信息检索数据融合线性组合法权重分配差分进化算法粒子群算法自适应交替粒子群差分进化算法　出处：《江苏大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着以网络技术为核心的现代信息技术的不断发展,如何帮助用户从互联网大量信息中迅速准确地获取用户需要的信息是信息检索的首要问题。数据融合技术能够将不同的检索系统所提交的检索结果进行组合从而得到一个新的检索结果。之前的的研究结果表明,数据融合技术能够有效地提高检索结果的性能。本文主要研究数据融合技术中的线性组合法,着重探讨如何采用智能优化算法解决线性组合法的权重分配问题。论文的主要工作如下:(1)本文探讨了基于差分进化算法和基于粒子群算法的权重分配策略。在上述两种优化算法的基础上,探讨了基于自适应交替的粒子群差分进化优化算法权重分配策略。该策略使用自适应的概率交替使用差分进化算法和粒子群算法对系统权重进行优化,可以有效避免粒子群算法已陷入局部极值的特点以及进一步强化差分进化算法的收敛能力。据我们所知,自适应交替粒子群差分进化算法是首次被应用到此类问题中。(2)为测试上述算法的有效性,采用TREC 2004 Robust Task数据集对不同数量的成员系统进行融合实验。实验结果表明,基于差分进化算法和基于粒子群算法的权重分配策略所得到的融合结果性能提升较为明显,而基于自适应交替差分进化粒子群优化算法的权重分配策略所得到的融合结果提升最为显著。(3)本文对所探讨的的三种融合方法在训练权重时所耗时间进行了比较。其中基于粒子群算法的权重分配策略训练耗时最少,基于自适应交替的粒子群差分进化优化算法权重分配策略次之,而基于差分进化算法的权重分配策略训练耗时最长。本文对已有的数据融合方法进行了简要的说明,从智能优化算法的角度提出了一种新的线性组合法权重分配策略,并通过实验比较这些融合方法的有效性和运行效率。实验结果表明,兼顾时间和性能,基于自适应交替粒子群差分进化优化算法权重分配策略能够有效地提升融合结果的性能。
[Abstract]:With the network technology as the core of the development of modern information technology. How to help users quickly and accurately obtain the information they need from a large amount of information on the Internet is the most important problem in information retrieval. Data fusion technology can combine the retrieval results submitted by different retrieval systems. Get a new search result. Previous research results show. Data fusion technology can effectively improve the performance of retrieval results. This paper mainly discusses how to use intelligent optimization algorithm to solve the weight assignment problem of linear combination method. The main work of this paper is as follows: 1). This paper discusses the weight allocation strategy based on differential evolution algorithm and particle swarm optimization algorithm. This paper discusses the weight allocation strategy of particle swarm optimization algorithm based on adaptive alternation, which uses adaptive probability alternately to optimize the weight of the system using differential evolution algorithm and particle swarm optimization algorithm. It can effectively avoid the characteristic that particle swarm optimization has fallen into local extremum and further enhance the convergence ability of differential evolution algorithm. Adaptive alternative Particle Swarm Optimization differential Evolution algorithm (APSO) is the first time to be applied to this kind of problem. TREC 2004 Robust Task dataset is used to perform fusion experiments on different number of member systems. The experimental results show that. The performance of the fusion algorithm based on differential evolution algorithm and particle swarm optimization algorithm is improved obviously. However, the fusion result obtained by the weight allocation strategy based on adaptive alternative differential evolution particle swarm optimization algorithm is the most significant. In this paper, we compare the time of the three fusion methods in training weight, and the training time of weight allocation strategy based on particle swarm optimization is the least. The weight allocation strategy of particle swarm optimization algorithm based on adaptive alternation is the second. The training time of weight allocation strategy based on differential evolution algorithm is the longest. In this paper, the existing data fusion methods are briefly explained. From the point of view of intelligent optimization algorithm, a new weight allocation strategy of linear combination method is proposed, and the effectiveness and efficiency of these fusion methods are compared by experiments. The experimental results show that both time and performance are taken into account. The weight allocation strategy based on adaptive alternative particle swarm optimization algorithm can effectively improve the performance of fusion results.
【学位授予单位】：江苏大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18;TP202

【参考文献】