高考志愿填报关键技术研究及系统实现

发布时间：2018-04-27 11:27

本文选题：高考志愿填报 + 院校投档线预测　；参考：《江苏大学》2017年硕士论文

【摘要】：高考志愿填报关乎考生一生的命运,然而每年都会有许多考生因为志愿填报失当导致“高分低就”甚至“高分不就”。高考志愿填报是一项庞大又复杂的系统工程。因为时间紧、信息多而导致家长及考生无法填出合理的志愿,所以开发一套志愿填报指导系统具有非常大的实际意义。对院校投档分数线的准确预测是志愿填报最为关键的问题。目前常用的预测方法主要有三种:两线差法、录取难度系数法以及分数排序定位法。这三种方法计算比较简单,但都是使用线性函数进行逼近,其预测结果与真实院校投档分数线存在较大误差。基于此,本文提出了基于神经网络的预测方法,并在此基础之上设计实现了一套高考志愿填报指导系统。本文在数据预测方面的工作主要包括两个方面:(1)基于神经网络的院校投档分数线的预测。针对往年录取数据与当年录取结果具有高度非线性的特点,本文提出一种基于神经网络实现院校投档分数线预测的方法。该方法首先使用粒子群算法对神经网络模型进行优化,通过粒子群算法在迭代过程中对粒子的自身历史能力认知和全局环境认知进行动态分配,解决了神经网络算法容易陷入局部极小值的问题。通过对江苏省2011-2015年的数据进行实验,实验结果表明与线差法相比,直接采用神经网络将本一、本二院校投档分预测误差不超过1分、2分的准确率分别提高了12%、14%,将本三、高职专科预测误差不超过3分的准确率提高了43%,这说明采用神经网络进行预测准确率更高。与直接使用神经网络进行相比,优化后的网络模型进一步将本一、本二院校投档分预测误差不超过1分、2分的准确率提高了4%、8%,将本三、高职专科预测误差不超过3分的准确率提高了4%,说明优化后的模型具有更高的预测准确率。(2)基于C4.5的六档专业推荐法。江苏省高考志愿填报规则规定每位考生同批次同科类填报8所院校,每所院校选择6个专业。针对现有基于线差法进行六档专业推荐的不足,首先提出了基于C4.5算法实现的六档专业推荐法,该方法在准确率方面比基于线差法专业推荐方法有了大幅提升。在此基础之上又采用了基于等价无穷小理论对C4.5选择分裂属性的进行优化,使得计算过程与直接采用原始C4.5算法相比所花费时间大大缩短。实验结果表明基于C4.5与基于线差法的六档专业推荐法相比,对本一、本二、本三以及高职专业批次在误差0档范围内分别提高了11%、17%、20%以及26%;而基于优化C4.5的六档专业分类法与基于C4.5六档专业推荐法相比,在保持精度不变的情况下,其时间复杂度由O(9)27)2)2n)变为O(9)2),效率方面得到明显提升。本文在系统实现方面的工作包括两个方面:(1)为了满足系统的业务需求以及后期的业务扩展,本文在数据库设计方面针对数据项太多的问题,提出数据表分表的方法;针对数据变动的问题,提出根据数据变动频率进行分表的方法;针对同一院校在同批次同科类下有多个招生代码的问题,提出不同招生代码对应院校做子父级关联的方法;针对高考政策经常性改革的问题,提出高考政策及志愿填报规则配置化管理方案。(2)为了提高系统对于Web请求的访问效率,本文提出一种基于Apache Shiro的Web高效访问控制方案,该方案采用对Web应用中的访问控制模块进行封装,并且形成滤器链的形式,允许链中任意相连模块进行通信,从而提升了系统的访问效率。实验结果表明,当系统在并发线程数达到5000时,采用该方案将吞吐量从24rps提高到41rps,提高了70%,将平均响应时间从1700ms降到1100ms,降低了35%。
[Abstract]:Voluntary filling in the college entrance examination is about the fate of the candidates' life. However, every year, many candidates are responsible for "high scores, low marks" and even "high scores". Voluntary filling in the college entrance examination is a huge and complex system project. Because of the tight time and information, the parents and candidates can not fill out a reasonable volunteer, so development A set of voluntary reporting guidance systems is of great practical significance. Accurate prediction of the grading line of colleges and universities is the most critical problem. There are three main methods of prediction: two line difference, admission difficulty coefficient and fractional ordering. These three methods are simple, but all are linear. In this paper, the prediction method based on neural network is proposed. Based on this, a set of voluntary reporting guidance system for college entrance examination is designed and implemented. The work of this paper mainly includes two aspects: (1) neural network based on Neural Network In this paper, a method based on neural network is proposed to predict the filing fraction line of colleges and Universities Based on neural network. Firstly, the method of particle swarm optimization is used to optimize the neural network model, and the particle swarm optimization algorithm is used in the iterative process. In order to solve the problem that the neural network algorithm is easy to fall into the local minimum, the problem that the neural network algorithm is easy to fall into the local minimum is solved by the cognition of its own historical ability and the global environment cognition. The experiment results of 2011-2015 years' data in Jiangsu province show that compared with the line difference method, the neural network directly uses the neural network to predict the error of the two colleges and universities. More than 1 points, the accuracy rate of 2 points increased by 12%, 14% respectively. The accuracy of this three, higher vocational college prediction error not exceeding 3 points was raised by 43%, which indicates that the accuracy rate of the neural network is higher. Compared with the direct use of neural network, the optimized network model will further this, and the prediction error of the 2 colleges and universities is not more than the prediction error. The accuracy of 1 points and 2 points is raised by 4%, 8%, and the accuracy of this three, higher vocational college prediction error not exceeding 3 points is raised by 4%, indicating that the optimized model has a higher prediction accuracy. (2) the C4.5 based six file professional recommendation method. The Jiangsu provincial college entrance examination voluntary reporting rules set each examinee to fill in 8 colleges and universities with the same classes. The school selects 6 specialties. Aiming at the shortage of the six grade professional recommendation based on the existing line difference method, this paper first puts forward the six file professional recommendation method based on the C4.5 algorithm. This method has been greatly improved in the accuracy rate than the line difference method professional recommendation method. On this basis, the C4.5 selection based on the equivalent infinitesimal theory is used. The optimization of the split attribute makes the calculation process shorter than the original C4.5 algorithm. Experimental results show that based on C4.5 and line difference based six professional recommendation method, this one, this two, this three, and higher professional batches have increased 11%, 17%, 20%, and 26% in the error range, respectively. Compared with the C4.5 six gear professional recommendation method, the optimization of C4.5's six gear classification method, with the constant precision, its time complexity changed from O (9) 27) to 2 2n) to O (9) 2), and improved the efficiency. The work of this paper in the system implementation includes two aspects: (1) to meet the business needs of the system and the later service. In order to solve the problem of data change, this paper puts forward the method of dividing tables according to the change frequency of data, and puts forward different enrolment codes corresponding to colleges and universities to be the parents for the same college in the same batch. In order to improve the access efficiency of the system for Web requests, a Web efficient access control scheme based on the Apache Shiro is proposed in order to improve the access efficiency of the system for Web requests. This scheme adopts the access control module in the Web application. Encapsulation, and form the form of a filter chain, allows the communication of any connected module in the chain to improve the system's access efficiency. The experimental results show that when the number of concurrent threads reaches 5000, the system improves the throughput from 24rps to 41rps, increases the throughput by 70%, reduces the average response time from 1700ms to 1100ms, and reduces the 35%.

【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】：TP311.52

【参考文献】