基于支持向量机的嫌疑人特征预测算法及分布式实现

发布时间：2018-03-26 01:29

本文选题：大数据　切入点：数据挖掘　出处：《合肥工业大学》2017年硕士论文

【摘要】：随着社会政治、经济和科技的高速发展,犯罪事件也以一定的速率不断增长,而且违法犯罪更具组织化、职业化和高智能化。我国公安信息系统信息化程度不高,分析研判不够智能化,决策机制有失科学性,缺乏对数据由宏观到微观的问题发现手段,如何利用数据挖掘的相关技术,充分发挥警务大数据的价值和作用,使其运用到警务工作中,提高执法效率和预防打击犯罪活动,已经成为公安信息化建设中急需解决的问题。因此本文针对大数据环境下,公安技术应用不足、备选嫌疑人众多而预测方法相对落后的问题,提出了运用支持向量机(SVM)预测犯罪嫌疑人的方法,提高侦破效率。传统的嫌疑人预测方法大都通过回归或者分类方法,对嫌疑人的可能性进行判断,这可能会导致错判的可能性。针对这一问题,本文对嫌疑人的特征进行预测,提出基于支持向量机的一种新颖的嫌疑人特征预测方法。首先,本文对支持向量机的基本原理进行介绍,在其基础上提出嫌疑人特征预测模型,并通过实验验证模型的有效性,针对大数据环境下嫌疑人特征预测问题,提出基于Hadoop的分布式嫌疑人特征预测框架。本文的研究成果主要有以下几个方面:(1)针对问题特性以及支持向量机的特点,将支持向量机算法运用到嫌疑人预测问题中。(2)提出嫌疑人特征预测模型。首先对数据进行预处理,并采用信息增益的特征选择方法进行特征选择,基于支持向量机构建嫌疑人特征预测模型,运用粒子群算法(PSO)对模型的参数进行优化,并通过实验对模型进行评估,验证其可行性。(3)提出基于Hadoop的分布式嫌疑人特征预测框架,解决海量数据嫌疑人特征预测问题。设计案件特征选择的并行化和分布式SVM的运行,并于单机的SVM进行对比实验分析,验证了Hadoop处理效率更高。本文的研究成果,不仅较好的解决了嫌疑人预测问题,也为嫌疑人预测、协助办案并提高办案效率提供了新的思路,具有一定的实际意义和借鉴价值。
[Abstract]:With the rapid development of social politics, economy and science and technology, the crime has been increasing at a certain rate, and the crime is more organized, professional and intelligent. Analysis and judgment is not intelligent, decision-making mechanism is not scientific, lack of data from macro to micro problem discovery means, how to make use of data mining related technology, give full play to the value and role of police big data, It has become an urgent problem in the information construction of public security to apply it to police work, to improve the efficiency of law enforcement and to prevent and crack down on criminal activities. Therefore, this paper aims at the insufficient application of public security technology in the environment of big data. In order to improve the detection efficiency, this paper puts forward the method of using support vector machine (SVM) to predict the criminal suspect, which is relatively backward in the prediction methods of many alternative suspects. Most of the traditional suspect prediction methods adopt regression or classification methods. Judging the possibility of suspect, this may lead to the possibility of misjudgment. In view of this problem, this paper predicts the feature of suspect, and puts forward a novel method of suspect feature prediction based on support vector machine. This paper introduces the basic principle of support vector machine, puts forward the suspect feature prediction model on the basis of it, and proves the validity of the model through experiments, aiming at the suspect feature prediction problem under big data environment. This paper proposes a framework for feature prediction of distributed suspects based on Hadoop. The main research results of this paper are as follows: 1) aiming at the characteristics of the problem and the characteristics of support vector machines, The support vector machine (SVM) algorithm is applied to the suspect prediction problem. (2) A suspect feature prediction model is proposed. Firstly, the data is preprocessed, and the feature selection method based on information gain is used to select the feature. Based on the support vector mechanism (SVM), the suspect feature prediction model is built, and the parameters of the model are optimized by particle swarm optimization (PSO), and the feasibility of the model is verified by experiments. (3) A distributed suspect feature prediction framework based on Hadoop is proposed. In order to solve the problem of suspect feature prediction in mass data, the parallelization of case feature selection and the operation of distributed SVM are designed, and compared with SVM on a single computer, it is verified that the efficiency of Hadoop processing is higher. It not only solves the problem of suspect prediction, but also provides a new way of thinking for suspect forecasting, assisting in handling cases and improving the efficiency of handling cases. It has certain practical significance and reference value.
【学位授予单位】：合肥工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：D917.6

【相似文献】