【摘要】：在創(chuàng)新型國(guó)家建設(shè)和國(guó)家科技體制改革的背景下,為了適應(yīng)國(guó)家科技發(fā)展趨勢(shì),高校對(duì)學(xué)生進(jìn)行科研能力培養(yǎng)與評(píng)價(jià)是十分必要的。傳統(tǒng)能力評(píng)價(jià)缺乏對(duì)評(píng)價(jià)指標(biāo)的重要度判斷,重要度判斷過(guò)程具有隨意性和主觀性,針對(duì)以上問(wèn)題,本文使用BP神經(jīng)網(wǎng)絡(luò)對(duì)大學(xué)生科研能力進(jìn)行評(píng)價(jià),旨在提高評(píng)價(jià)過(guò)程的科學(xué)性和精確性,完善大學(xué)生培養(yǎng)過(guò)程,為高校了解學(xué)生科研能力提供較為科學(xué)的依據(jù)。本文首先基于大學(xué)生科研能力培養(yǎng)的過(guò)程、論文寫(xiě)作流程,分析每個(gè)過(guò)程中涉及到的能力,構(gòu)建了大學(xué)生科研能力評(píng)價(jià)指標(biāo)體系;綜合利用形成性評(píng)價(jià)和終結(jié)性評(píng)價(jià)的方式,設(shè)計(jì)了大學(xué)生科研能力指標(biāo)評(píng)分細(xì)則;根據(jù)評(píng)分細(xì)則表設(shè)計(jì)了網(wǎng)絡(luò)調(diào)查問(wèn)卷,根據(jù)收集的問(wèn)卷數(shù)據(jù)基于組合賦權(quán)法對(duì)指標(biāo)賦權(quán)。然后針對(duì)評(píng)價(jià)指標(biāo)的非線性特征,達(dá)到弱化傳統(tǒng)評(píng)價(jià)方法存在的隨機(jī)性和主觀性的效果,實(shí)現(xiàn)能力評(píng)價(jià)的科學(xué)性和實(shí)用性,構(gòu)建了基于BP神經(jīng)網(wǎng)絡(luò)的評(píng)價(jià)模型;為了加強(qiáng)訓(xùn)練樣本的可比性,利用min-max方法對(duì)樣本內(nèi)容進(jìn)行標(biāo)準(zhǔn)化處理;根據(jù)樣本的選擇規(guī)則,從問(wèn)卷數(shù)據(jù)中選取了等量的樣本用來(lái)訓(xùn)練和測(cè)試;利用試湊法確定了隱層神經(jīng)元的數(shù)量;應(yīng)用選取的樣本對(duì)基于梯度下降法、擬牛頓法、列文伯格法(Levenberg-Marquardt,LM)的BP神經(jīng)網(wǎng)絡(luò)在matlab軟件中進(jìn)行實(shí)驗(yàn),從網(wǎng)絡(luò)均方誤差、迭代次數(shù)、泛化能力、預(yù)測(cè)準(zhǔn)確率等四個(gè)評(píng)價(jià)指標(biāo)對(duì)3種BP神經(jīng)網(wǎng)絡(luò)模型針對(duì)本問(wèn)題進(jìn)行了可行性驗(yàn)證,實(shí)驗(yàn)結(jié)果表明基于LM算法的8-12-1的單隱層BP神經(jīng)網(wǎng)絡(luò)評(píng)價(jià)模型應(yīng)用于大學(xué)生科研能力評(píng)價(jià)是可行的。最后對(duì)原型系統(tǒng)進(jìn)行設(shè)計(jì)與實(shí)現(xiàn),從需求分析、數(shù)據(jù)庫(kù)設(shè)計(jì)、核心模塊三個(gè)方面進(jìn)行了闡述,設(shè)計(jì)并實(shí)現(xiàn)了 BP神經(jīng)網(wǎng)絡(luò)模型管理模塊和能力評(píng)價(jià)模塊,初步實(shí)現(xiàn)了評(píng)價(jià)大學(xué)生科研能力的原型系統(tǒng),為大學(xué)生科研管理提供了實(shí)用的工具。
[Abstract]:Under the background of the construction of innovative country and the reform of national science and technology system, it is very necessary for colleges and universities to cultivate and evaluate students' scientific research ability in order to adapt to the development trend of national science and technology. The traditional ability evaluation lacks the importance judgment to the evaluation index, the important degree judgment process has the arbitrariness and the subjectivity, in view of the above question, this article uses the BP neural network to carry on the appraisal to the university student scientific research ability. The purpose of this paper is to improve the scientificity and accuracy of the evaluation process, to perfect the cultivation process of college students, and to provide a scientific basis for the understanding of students' scientific research ability in colleges and universities. Firstly, based on the process of cultivating college students' scientific research ability, the thesis writing process, analyzing the ability involved in each process, constructing the evaluation index system of university students' scientific research ability, synthetically utilizing the formative evaluation and summative evaluation methods. The evaluation rules of scientific research ability index of college students are designed, and the network questionnaire is designed according to the scoring rules table, and the index is weighted according to the collected questionnaire data based on the combination weight method. Then aiming at the nonlinear characteristics of the evaluation index, the evaluation model based on BP neural network is constructed by weakening the randomness and subjectivity of the traditional evaluation method and realizing the scientificity and practicability of the ability evaluation. In order to enhance the comparability of training samples, the min-max method is used to standardize the contents of the samples, and according to the selection rules of the samples, the same number of samples are selected from the questionnaire data for training and testing. The number of hidden layer neurons is determined by trial and error method, and the BP neural network based on gradient descent method, quasi-Newton method and Levenberg-Marquardt LM method is used to test the BP neural network in matlab software. Four evaluation indexes, such as generalization ability and prediction accuracy, are used to verify the feasibility of three BP neural network models. The experimental results show that the application of 8-12-1 single hidden layer BP neural network evaluation model based on LM algorithm to the evaluation of university students' scientific research ability is feasible. Finally, the prototype system is designed and implemented, including requirement analysis, database design and core module. The model management module and capability evaluation module of BP neural network are designed and implemented. The prototype system for evaluating the scientific research ability of college students is implemented preliminarily, which provides a practical tool for the management of university students' scientific research.
【學(xué)位授予單位】：大連海事大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2017
【分類(lèi)號(hào)】：G642;TP183

【參考文獻(xiàn)】

相關(guān)期刊論文 前10條

1 劉春霞;田蕓;;高�？蒲心芰Φ膮f(xié)同IWD粗糙集-塊神經(jīng)網(wǎng)絡(luò)評(píng)估模型[J];計(jì)算機(jī)工程與科學(xué);2016年03期

2 鄭金山;郎奠波;;高校教師科研能力評(píng)估模型研究[J];哈爾濱師范大學(xué)自然科學(xué)學(xué)報(bào);2016年02期

3 王亮軍;李國(guó)寧;劉雨佳;;基于粒子群聚類(lèi)算法的模糊神經(jīng)網(wǎng)絡(luò)建模方法研究[J];微電子學(xué)與計(jì)算機(jī);2016年02期

4 宋曉勇;陳年生;;遺傳算法和神經(jīng)網(wǎng)絡(luò)耦合的金融預(yù)測(cè)系統(tǒng)[J];上海交通大學(xué)學(xué)報(bào);2016年02期

5 王林;彭璐;夏德;曾奕;;自適應(yīng)差分進(jìn)化算法優(yōu)化BP神經(jīng)網(wǎng)絡(luò)的時(shí)間序列預(yù)測(cè)[J];計(jì)算機(jī)工程與科學(xué);2015年12期

6 于勁松;沈琳;唐荻音;劉浩;;基于貝葉斯網(wǎng)絡(luò)的故障診斷系統(tǒng)性能評(píng)價(jià)[J];北京航空航天大學(xué)學(xué)報(bào);2016年01期

7 冷昕;張樹(shù)群;雷兆宜;;改進(jìn)的人工蜂群算法在神經(jīng)網(wǎng)絡(luò)中的應(yīng)用[J];計(jì)算機(jī)工程與應(yīng)用;2016年11期

8 朱劍;;科研體制與學(xué)術(shù)評(píng)價(jià)之關(guān)系——從“學(xué)術(shù)亂象”根源問(wèn)題說(shuō)起[J];清華大學(xué)學(xué)報(bào)(哲學(xué)社會(huì)科學(xué)版);2015年01期

9 黃宇;李戰(zhàn)國(guó);馮愛(ài)明;;高校科研創(chuàng)新團(tuán)隊(duì)建設(shè):困境與突圍[J];高等工程教育研究;2013年02期

10 徐恪;朱敏;林闖;;互聯(lián)網(wǎng)體系結(jié)構(gòu)評(píng)估模型、機(jī)制及方法研究綜述[J];計(jì)算機(jī)學(xué)報(bào);2012年10期

相關(guān)博士學(xué)位論文 前1條

1 鞠儒生;基于數(shù)據(jù)耕種與數(shù)據(jù)挖掘的系統(tǒng)效能評(píng)估方法研究[D];國(guó)防科學(xué)技術(shù)大學(xué);2006年

，

本文編號(hào)：2120055