【作者】 许敏；

【导师】 王士同；

【作者基本信息】 江南大学 ， 轻工信息技术与工程， 2009， 硕士

【摘要】 高职教育大众化使得高职院校办学规模不断扩大,导致师资力量相对紧缺,致使教学质量问题日渐明显。针对这一问题,教育部组织有关专家分批对各高职院校进行评估。因此,如何提高教学质量已经成为高职院校当前和今后办学的首要任务,要提高教学质量,必须加强对教学质量的全面管理,尤其是对教师教学质量的评价。对教学质量的评价应以一定的教育目标为标准,运用有效可行的技术手段,对教育教学活动的过程、方法、形式、成绩等进行价值判断、评估,从而形成正确的、稳定的价值判断标准和氛围,使教师产生信任感、公平感和成就感,评价后运用信息反馈,还能使教师及时获得教学过程中的各种信息,看到自己的成绩与不足,及时强化、调节、矫正,不断克服教学中的失误和不足,发扬成绩,使教学水平更上一层楼,但如何建立科学合理的教学质量评价体系是一个比较复杂的问题。就现有的研究状况来看,集中在两个方面,一是如何最有效、方便的收集到测评数据;二是如何对这些数据进行处理,达到公平评定教学质量等级的目的。教学质量评价时,影响教学质量的因素很多,且各因素的影响程度也不同,故评价结果难以用恰当的数学解析式表示,属于复杂的非线性分类问题,传统的分类方法不能很好地解决这些问题。人工神经网络作为一种新技术,以其非线性映射、学习分类和实时优化等基本特性为模式识别、非线性分类等研究开辟了新途径,在各类评价问题上应用很多。但也存在一些缺陷:容易陷入局部极小,网络收敛速度较慢,泛化能力较弱或很差。因此本课题在此基础上提出了一种将粒子群优化算法(PSO)训练的神经网络用于高职院校教师教学质量综合评价的方法。将PSO与BP神经网络结合,用PSO算法来优化神经网络的连接权值与阈值,可以较好的克服BP神经网络的固有缺陷,与BP算法比较,该方法在提高误差精度的同时可以加快训练收敛的速度,其泛化性能也比较好。本课题立足于高职院校教学工作的实际特点和教育发展对教学质量的要求,以学校内部人才培养质量评价体系为研究对象,将PSO优化的人工神经网络理论引入高校教学质量评价中。采用BP神经网络模型结构建立数学模型,详细介绍了PSO优化的BP神经网络算法。本系统的实现分成两个部分,一是通过网上测评系统收集学生、教师、专家的打分数据;二是通过评价系统对收集到的数据进行处理,实现样本维护、PSO优化的神经网络训练以及神经网络评价等功能,以获得教师的最终评价成绩。通过实验数据表明,PSO优化神经网络用于教学质量评价完全可行,而且满足精度要求,为高职院校教学质量评价提供了方便实用的工具。更多还原

【Abstract】 As the higher vocational education become more and more popular, it caused not only the higher educational scale is becoming larger, but also the shortage of teacher resource, the teaching quality problem is becoming obvious. Against the problem, the department of education organized some experts to assess every higher college. So, how to improve the teaching quality has already been the first task to every higher vocational education in the recent time and future. In order to improve the quality, we must strengthen the management whole and strict, especially on the evaluation of teaching quality. The quality evaluation should be on some educational objective for expenses, use the useful and feasible technical method course to justify and evaluate the educational activity’s procedure, method, form and goal etc, the result is to form the correct steady value standard and atmosphere, make the teachers have the feelings of convince fair and achievement. After the evaluation, using the feedback, the teachers can achieve every information in the educational process in time, know their own achievements and shortage, strengthen adjust remedy constantly overcome the misplay and shortage in education in time, develop achievements, make the level of education better and better. But how to set up a scientific justice quality system is a complex problem. On the existing situation of study, it concentrates on two points: one is how to get the data conveniently and effectively; another is on the study of the way how to assess quality grade.In the quality evaluation process, as there are many factors affecting it, the degree they are affected is different and the result of evaluation is hardly to use the equal mathematics analytic expression to show. It is a complicated non-linear classification problem. Traditional classification method can’t solve these problems thoroughly. BP network is a new kind technique, on their non-linear mapped, studying sort and real time optimization etc. These characters cut a new way for the study of mode identify, non-linear mapped and widely used in many kinds of evaluation problems. BP algorithm has some shortcomings such as the speed of training is slow and the generation probably is not good. So, this article uses PSO algorithm to train BP network in order to get the evaluation of the teaching quality. It suggests that this algorithm can reduce number of training and error obviously and has better generalization than traditional BP algoritlim.The subject is established in the characters of the higher vacation educational work and the request of educational quality to educational development, using the system of bringing up person with ability as the studying objective, bringing the BP neural network trained by PSO algoritlim theory into the higher quality evaluation. Using BP network model structure to set up math model, introduce the arithmetic BP neural network trained by PSO algorithm.The system’s realization is divided into two parts: one is to get the data from the evaluation system through network; another is to deal with the data, such as the functions of stylebook maintenance, BP neural network which trained by PSO algorithm’s training and evaluation. Through the experiment data, it indicates using BP neural network which trained by PSO algorithm evaluate quality is feasible, and satisfy the request of precision, it really prompts a convenient tool for higher education evaluation.更多还原