The Design of College Student Achievement Management System based on GA-BP Network

doi:10.21203/rs.3.rs-2462027/v1

Download PDF

Research Article

The Design of College Student Achievement Management System based on GA-BP Network

https://doi.org/10.21203/rs.3.rs-2462027/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 04 Apr, 2023

Read the published version in Soft Computing →

You are reading this latest preprint version

With the development of computer technology, higher education and teaching work gradually depend on the score management system. The performance management system has a substantial amount of student curriculum performance data, and how to effectively analyze these data to improve the teaching quality of the school is a problem worthy of in-depth study. This study integrates a genetic algorithm and a neural network on the basis of the BP neural network to create a GA-BP network hybrid algorithm that addresses the BP neural network's issues, such as its proclivity to fall into local minima and slow convergence speed. The GA-BP network hybrid algorithm improves the connection weights and thresholds of the BP network using the genetic algorithms global search capability so that the network can search from a better initial value and complete the global optimization more quickly. According to the needs of student achievement management, this research designed and developed a student achievement management system and realized the application of the GA-BP network in the student achievement management system. Through the experiments of MATLAB software simulation, this study has verified the feasibility of the model applied to the graduation score prediction. The results show that the model can make a more accurate prediction of the unknown graduation scores by using the existing student course scores, so as to use the prediction results to carry out academic warning prompts for students, help students strengthen their professional knowledge learning, and help colleges and universities adjust their education and teaching work.

Performance prediction

data mining

GA-BP network

Genetic Algorithm

At present, data mining technology is developing rapidly and has been successfully applied in e-commerce, marketing, traffic management, social governance and other fields ^[1]. According to the definition of data mining proposed by Fayy et al., data mining is the process of finding effective, valuable, potential and easily understood data from data sets ^[2]. Data mining is an interdisciplinary and comprehensive emerging discipline, mainly involving database, mathematical statistics, data warehouse, artificial intelligence and other fields ^[3].

Education data mining is a process of using data mining technology to study a large number of existing data related to students stored in the school educational administration management system and extract information and knowledge with practical value to improve students' knowledge level and improve school education and teaching quality ^[4]. At present, the analysis of students' course performance data is only some simple operations, such as calculating the average score, pass rate, class ranking, professional ranking, etc., but these operations cannot find the internal relationship between grades. For example, how to use the existing students' course scores to predict the unknown test scores, how the students' course scores affect their graduation scores, and which courses have played an important role are all questions worthy of research.

In view of the above problems, many scholars have applied data mining technology in the field of education and made some achievements. For example, Romero studied how to use different data mining technologies to improve curriculum and students' learning ^[5]. He also compared traditional classroom teaching with online education system ^[6]. Shaeela uses data mining technology to improve the school's teaching system, which has significantly improved the school's teaching level ^[7]. Rashan discussed how neural networks can help teachers better manage their classrooms. Tisela uses data mining technology to predict closely related topics in the curriculum and further improve the curriculum of educational institutions ^[8].

However, the student achievement management system is still in the stage of spontaneous construction, mainly focusing on expository information construction (such as timetable arrangement, teaching plan, teacher information, curriculum teaching plan information, etc.), and there are problems such as non-standard and inconsistent system format and content, strong randomness of frame design, and there is no systematic, dynamic and real-time integration of information within the system ^[10]. At the same time, these studies focus on real-time monitoring of the entire teaching process, and there is a problem that students do not pay attention to the process of statistics, input and analysis of student performance data ^[11].

In view of the above problems, this research proposes a hybrid algorithm of GA-BP network. This model can make more accurate predictions of the unknown graduation scores by using the existing students' curriculum scores. It has certain reference value for determining training plans and curriculum settings, and also has a good guiding significance for teachers' teaching work and students' learning.

As the quality of information construction in colleges and universities continues to advance, Internet information technology is gradually changing the traditional mode of education and teaching management, making the work efficiency of educators continuously improve. Among them, the design and implementation of college student achievement management system has become the research direction of many scholars.

Europe and the United States have always attached great importance to the construction of education informatization, and many experienced academic teams have provided technical services and support for the construction of education informatization ^[12]. In 1973, the Massachusetts Institute of Technology first put forward the slogan of "digital campus construction"^[13], and then many countries and universities joined in the process of digital campus construction. Up to now, more than 90% of schools in Europe and the United States have adopted information technology in teaching management ^[14]. The student accomplishment management system and the student achievement management system are commonly used in American colleges and universities and the educational administration management system have formed 100% information sharing ^[15]. In Japan, student achievement is an important criterion for Japanese universities to evaluate students. The student achievement management system realizes the sharing and interaction of student achievement information, and ensures the accuracy and security of information ^[16].

Data mining technology is to make statistics and analysis on the massive data in the database, find rules and connections from them, and make predictions, thus giving guidance to people ^[17]. At present, the commonly used methods in data mining technology include association rules, decision trees, neural networks, clustering analysis and rough sets, which generally include the steps of determining mining objects, data preparation, data mining, result analysis, etc. The specific mining process is shown in Fig. 1.

Applying data mining technology to education is a research hotspot in recent years. The application of data mining technology in the field of education is a process of extracting information and knowledge with practical value by analyzing a large number of data related to students stored in the school educational administration management system, so as to improve the level of students' knowledge and improve the quality of school education^[18]. The classification algorithm proposed by professor J.R. Quinlan of Australia and the association algorithm proposed by Agrawal of international business Machine Company are outstanding achievements in this field, and have promoted the emergence of a number of excellent data mining software applications ^[19]. Kirimi made use of association rules to mine semester subject scores, trying to find hidden factors that affect discipline development, as well as the relationship between various factors that affect discipline development ^[18]. Ahmed applied data mining technology to the application research of network teaching system, and created a framework model of network teaching system to guide students' learning and teachers' teaching^[19]. Qin discussed the feasibility of data mining on the selection of sports talents, excavated a large amount of accumulated data, looked for the interrelationship of various factors, and found some new laws of knowing school sports scientific research and teaching training ^[20].

This study selected the BP neural network method to satisfy the performance management needs of college students. We propose a hybrid algorithm of GA-BP network that combines genetic algorithms with neural networks to address the issue that BP neural networks are prone to falling into local minima and have sluggish convergence rates.

Due to the capabilities of associative memory, categorization, and self-organization that artificial neural networks possess; it has good effect in dealing with nonlinear problems. In recent years, neural networks have made great progress in predicting students' grades, therefore, we employ BP neural network to investigate how college students' marks are managed.

3.1 The establishment of BP neural network model

The neural network's input in this study is made up of 12-dimensional student achievement data. The network's input layer has 12 nodes, its hidden layer maps any nonlinear functions, and its output layer uses one node to represent student achievement. Figure 2 depicts the grade model for prediction based on the BP neural network.

However, there are frequently issues with slipping into local minimum points and delayed convergence when employing a BP neural network to solve the accomplishment prediction of college students. The GA-BP network is created as a result of this research's optimization of the BP neural network by integrating genetic algorithms with neural networks. The approach makes use of the genetic algorithm's global search capability to increase the BP network's link weights and thresholds so that the network searches from a better starting value and overcomes its flaws.

3.2 Establishment of GA-BP Neural Network Model

3.2.1 Choice of encoding

The commonly used coding methods are binary coding and real number coding, but this research belongs to the optimization problem with high dimensionality. Using binary coding will cause too many chromosome genes, and it needs to be decoded after coding, which is a cumbersome process. Therefore, this research chooses real number coding to reduce the computational complexity, improve the efficiency of the operation, and avoid the mapping errors in binary coding^[21]. The connection weight between the input layer and the hidden layer, the hidden layer threshold, the connection weight between the hidden layer and the output layer, and the output layer threshold make up each of the four components of a BP neural network chromosome. In the BP neural network, each connection weight and threshold are represented by a real number.

Assuming that each chromosome in the population contains $S$ genes, there are:

$$S=n\times m+m\times l+m+l \left(1\right)$$

The input layer, hidden layer and output layer have corresponding number of nodes ($n, m$ and $l$) respectively.

Therefore, for the three-layer BP neural network designed, the number of chromosome codes is $S=12\times 4+12\times 1+4+1=65$, that is, a total of 65 parameters needs to be optimized.

3.2.2 Design of fitness function

Only fitness function is used by genetic algorithms as a criterion for data optimization and evaluation. As a result, the overall effectiveness of the genetic algorithm is closely tied to how the fitness function is designed. Since the fitness function determines the likelihood of survival, its value must be positive, which encourages competition between individuals.

Typically, the optimized goal function serves as the foundation for the fitness value function design. The mean square error function produced by the network serves as the evolutionary algorithm's fitness value function in this study. The mean square error function is shown in formula (2).

$$MSE=\frac{1}{mp}\sum _{p=1}^{p}\sum _{j=1}^{m}{({\widehat{y}}_{pj}-{y}_{pj})}^{2} \left(2\right)$$

Since the genetic algorithm selects individuals with large fitness function values, the fitness value function takes the reciprocal of the mean square error.

$$f=\frac{1}{MSE} \left(3\right)$$

In this way, an individual with a large fitness value is an individual with a small mean square error function, so that the population can evolve towards a better value.

3.2.3 Genetic manipulation

(1) Choose the best individual

The goal of selection is to choose better people so that they can pass on their traits to the next generation either directly or indirectly through crossover and mutation processes. The individual's fitness is taken into account when choosing the selection procedure. The likelihood of selection increases with fitness level. On the contrary, the individual with lower fitness will be eliminated.

We adopt the roulette selection method, and the relative fitness is defined as:

$$p\left({x}_{i}\right)=\frac{f\left({x}_{i}\right)}{\sum _{j=1}^{N}f\left({x}_{i}\right)} \left(4\right)$$

Among them, $p\left({x}_{i}\right)$ is the relative fitness of individual ${x}_{i}$, that is, the probability of individual ${x}_{i}$ being selected; $f\left({x}_{i}\right)$ is the original fitness of individual ${x}_{i}$. $\sum _{j=1}^{N}f\left({x}_{i}\right)$ is the cumulative fitness of the population.

Divide a disk into corresponding N sectors according to the selection probability $p\left({x}_{i}\right)$ of each individual, where the central angle of the $i$-th sector is:

$$2\pi \frac{f\left({x}_{i}\right)}{\sum _{j=1}^{N}f\left({x}_{i}\right)}=2\pi p\left({x}_{i}\right) \left(5\right)$$

Then set a fixed pointer. When making a selection, you can imagine rotating the roulette. If the pointer points to the $i$-th sector after the roulette is stationary, select the individual $i$, as shown in Fig. 3.

(2) Real number cross method

The crossover between the $i$-th chromosome of chromosome ${a}_{i}$ and the $j$-th chromosome of chromosome ${a}_{j}$ at the r position is done using the real number crossover method.

$$\begin{array}{c}{a}_{ir}={a}_{ir}\left(1-c\right)+{a}_{jr}c\\ {a}_{jr}={a}_{jr}\left(1-c\right)+{a}_{ir}c\end{array} \left(6\right)$$

Here, $c$ is a random number between [0, 1].

(3) Genetic mutation manipulation

${a}_{ij}=\left\{\begin{array}{c}{a}_{ij}+({a}_{ij}-{a}_{max}){r}_{2}{\left(1-\frac{g}{{G}_{max}}\right)}^{2} {r}_{1}\ge 0.5\\ {a}_{ij}+\left({a}_{min}-{a}_{ij}\right){r}_{2}{\left(1-\frac{g}{{G}_{max}}\right)}^{2} {r}_{1}<0.5\end{array}\right.$ (7)

Here, ${a}_{max}$ and ${a}_{min}$ are the gene ${a}_{ij}{\prime }s$ upper and lower limits, respectively; ${r}_{1}$ is a random number between [0, 1]; ${r}_{2}$ is a random number; and $g$ is the current iteration number. The maximum number of evolutions is ${G}_{max}$.

3.2.4 Stop condition

The algorithm has three termination conditions, the first of which is when the fitness of the ideal person reaches a certain level. Second, a predetermined number of algebras are reached by the number of iterations. Thirdly, neither the fitness of the ideal person nor the fitness of the group continues to rise.

When at least one of the aforementioned conditions is met, the algorithm ends. In this study, the algorithm ends after the iteration reaches the predetermined algebra, and one of 100 default generations is selected.

3.3 GA-BP network hybrid algorithm flow

The algorithm flow is described as follows:

(1) The weights and thresholds of the model are real-number coded, and an initial population $W={({W}_{1},{W}_{2},\dots {W}_{100})}^{T}$ with 100 individuals is produced at random.

(2) Through the fitness function, which uses the inverse of the mean square error function of the model's actual output and expected output, each individual's proper value is determined.

(3) Use the roulette method to select, select some individuals with high fitness value as the parent, and discard the individuals with small fitness value.

(4) Two genetic operators that can be employed to modify the parent to produce children are crossover and mutation. The parent will be replaced by the offspring as the parent if the offspring has a higher fitness value than the parent. The parent generation consists of the same number of individuals.

(5) The stopping condition is attained by repeating steps (2) through (4) and carrying out a new cycle of crossover, mutation, and selection processes on the newly created individuals.

(6) By selecting the best member of the present group, the value on the chromosome is divided into the connection weight and threshold of the model, and the best weight and threshold are used as the BP neural network's initial values.

(7) With pre-set parameters, train the network model.

(8) When the training objectives are met, the training is terminated.

Figure 4 depicts the flowchart for the genetic algorithm optimization of the network model.

3.4 Design of student achievement management platform based on GA-BP network model

We design a student achievement management system based on the GA-BP neural network prediction model, and plan the required functional modules in detail. The student achievement management system includes the following features:

(1) Data maintenance module: data maintenance business is mainly used to maintain the information of students, teachers, classes and corresponding courses. Although these businesses are simple in operation and implementation, the main businesses involved are just adding, deleting, modifying and querying. But these data affect the stability and security of the system.

(2) Examination management module: As the core module of this system, examination management mainly includes the sub modules of examination organization, examination paper management, examination paper marking management and anti-cheating management. The business subject of this part is teachers, and the organization of examination business is the processing of examination information, including examination class, examination time, examination place, etc. The marking management is that the teacher manually or automatically marks the students' papers after each examination; The data involved in the examination management system is the core data of the system.

(3) Score prediction module: the score prediction module is also the core function module of this system. This module's primary purpose is to forecast students' graduation results using the GA-BP neural network that is suggested, and give early warning to students in danger.

(4) After class learning module: After class learning includes self-test management module and class assignment module. Students can use self-test management to learn after class, and the system can also remind students of self-study. In addition, lecturer users assign homework through class assignments, and student users view and submit homework through class assignments. The system functional structure design is shown in Fig. 5.

After obtaining the optimized GA-BP neural network model, we apply it to the college student achievement management system, and verify the prediction accuracy and performance of the model to solve the problem of achievement prediction through simulation experiments.

4.1 Dimension reduction of sample data

The data set of this study is the achievement data of the 4th year graduates from the College of Electronic Engineering in a university. Firstly, PCA dimension reduction is performed on the sample data. The following results can be obtained after the analysis of 12-dimensional student achievement data samples by SPSS software:

(1) Table 1 shows that the resulting 12 primary components' contribution rates are, in descending order, 51.585%, 10.634%, 8.019%, 6.333%, 5.148%, 4.238%, and 1.214%. As a result, the first six major components have a cumulative contribution rate of 85.96%. At the same time, the eigenvalues of this study are set to be greater than 0.5, and the eigenvalues of these six principal components meet the requirements. Therefore, it is appropriate to select these six principal components instead of the 12 indexes of the original sample as the input variables of the model.

Table 1
Total Variance Explained
	Initial eigenvalue			Extract the Sums of Squared Loadings			Rotation Sums of Squared Loadings
	Sum	Variance Proportion	Cumulative Percent	Sum	Variance Proportion	Cumulative Percent	Sum	Variance Proportion	Cumulative Percent
1 2 3 4 5 6 7 8 9 10 11 12	6.190 1.276 .962 .760 .618 .509 .430 .388 .316 .224 .182 .146	51.585 10.643 8.019 8.333 5.148 4.238 3.586 3.323 2.629 1.867 1.516 1.214	51.585 62.219 70.237 76.570 81.718 85.956 89.542 92.774 95.403 97.271 98.786 100.000	6.190 1.276 .962 .760 .618 .509	51.585 10.643 8.019 6.333 5.148 4.238	51.585 62.219 70.237 76.570 81.718 85.956	2.944 2.884 1.203 1.166 1.067 1.051	24.532 24.037 10.027 9.715 8.888 8.757	24.532 48.569 58.596 68.310 77.199 85.956

(2) Table 2 is the component matrix of data samples, which is the basis for obtaining the principal component score matrix.

Table 2
Component matrix
	Assembly
	1	2	3	4	5	6
College English Higher Mathematics Signal and system Digital circuit Analog Electronic Technology Algorithm and data structure Engineering Mathematics College Physics Fundamentals of computer Programming Circuit analysis principle	.875 .853 .847 .843 .820 .799 .758 .709 .132 .557 .602	− .027 − .120 − .016 − .082 − .209 − .098 − .246 − .074 − .727 − .659 − .133	− .071 − .005 − .053 − .007 − .062 − .103 − .099 − .153 − .586 − .670 − .304	− .118 − .004 − .083 − .113 − .205 − .023 − .086 − .341 − .058 − .147 − .710	.026 .029 − .107 .158 .192 − .128 − .023 .232 .267 .177 .032	− .325 .025 .160 − .026 .197 − .362 − .170 .234 − .132 .160 .213

Extraction method: principal component analysis

(3) The major component score coefficient matrix is shown in Table 3. This table gives the specific scores of the six principal components. According to these scores, the specific value of each principal component can be calculated. Then, according to the component score coefficient matrix, there are:

$${PC}_{1}=-0.171{X}_{1}+0.581{X}_{2}+0.357{X}_{3}+0.031{X}_{4}-0.106{X}_{5}+0.636{X}_{6}-0.192{X}_{7}+0.088{X}_{8}-0.099{X}_{9}-0.366{X}_{10}+0.052{X}_{11}-0.238{X}_{12}$$

$${PC}_{2}=-1.310{X}_{1}-0.243{X}_{2}-0.176{X}_{3}+0.016{X}_{4}+0.043{X}_{5}-0.291{X}_{6}-0.113{X}_{7}+0.246{X}_{8} +0.107{X}_{9}+0.651{X}_{10}+0.149{X}_{11}+0.532{X}_{12}$$

$${PC}_{3}=-0.065{X}_{1}-0.054{X}_{2}-0.026{X}_{3}-0.018{X}_{4}+0.030{X}_{5}-0.248{X}_{6}+1.055{X}_{7}-0.092{X}_{8} +0.215{X}_{9}-0.132{X}_{10}+0.060{X}_{11}-0.057{X}_{12}$$

$${PC}_{4}=1.104{X}_{1}-0.227{X}_{2}+0.001{X}_{3}-0.132{X}_{4}-0.080{X}_{5}-0.038{X}_{6}-0.058{X}_{7}-0.179{X}_{8} +0.241{X}_{9}-0.049{X}_{10}-0.016{X}_{11}-0.114{X}_{12}$$

$${PC}_{5}=-0.084{X}_{1}-0.057{X}_{2}+0.137{X}_{3}-0.056{X}_{4}+0.973{X}_{5}-0.174{X}_{6}+0.011{X}_{7}-0.006{X}_{8} +0.096{X}_{9}-0.032{X}_{10}-0.023{X}_{11}+0.063{X}_{12}$$

$${PC}_{6}=-0.086{X}_{1}+0.071{X}_{2}+0.114{X}_{3}+0.967{X}_{4}-0.056{X}_{5}-0.046{X}_{6}+0.005{X}_{7}+0.067{X}_{8} -0.167{X}_{9}+0.138{X}_{10}-0.055{X}_{11}-0.092{X}_{12}$$

Table 3
Component score coefficient matrix
	Assembly
	1	2	3	4	5	6
College English Higher Mathematics Signal and system Digital circuit Analog Electronic Technology Algorithm and data structure Engineering Mathematics College Physics Fundamentals of computer Programming Circuit analysis principle	− .171 .581 .357 .031 − .106 .636 − .192 .088 − .099 − .366 .052	− .131 − .243 − .176 .016 .043 − .291 − .113 .246 .107 .651 .149	− .065 − .054 − .026 − .018 .030 − .248 1.055 − .092 .251 − .132 .060	1.104 − .223 .001 − .132 − .080 − .038 − .179 .234 − .038 − .033 .076	− .084 − .057 .137 − .056 .973 − .183 .022 − .032 .098 − .023 − .063	− .098 .042 .113 .987 − .032 − .024 .004 .023 .023 − .023 − .034

Extraction method: principal component analysis

To sum up, after using the principal component analysis, the six principal components selected according to the cumulative contribution rate and eigenvalue can contain most of the information of the sample data, and it is appropriate and effective to replace the original input variables of the BP neural network.

4.2 Application of GA-BP neural networks in students' performance prediction

Through principal component analysis, we obtained 2000 groups of 8-dimensional student performance data samples, including 1200 groups of data as training samples, 300 groups of data as validation samples, and 500 groups of data as test samples. We use validation samples to determine the structural attributes and parameter values of the model. We used 1200 training samples to train GA-BP prediction model.

The experimental results show that after about 20 iterations of the prediction model, the model training error is less than 0.012, which proves that the model has high application value.

The fact that the two curves in Fig. 7 are essentially identical shows that the model we created can successfully predict students' graduation results.

Table 4

Training fitting error and prediction error of GA-BP prediction model
Standard	Training sample fitting	Test sample prediction
Mean square error(MSE)	2.1741	2.3504
Average relative error(MAPE)	0.0162	0.0165

The average relative error produced by the model during training is 0.0162, and the average relative error produced during prediction is 0.0165, as shown in Table 4. This demonstrates that the prediction model we developed has a high prediction accuracy during the training and prediction processes and has a good application value in forecasting student graduation performance situations.

This research takes the field of data mining as its background, uses the GA-BP network to predict the corresponding graduation grades according to the students' course grades and other data, and uses the MATLAB software to simulate the prediction model so as to achieve the goal of students' academic early warning. The results achieved are as follows:

On the basis of the network model, the genetic algorithm and the neural network are creatively combined to develop a GA-BP network hybrid algorithm, which addresses the problems with the network's inclination to fall into local minima and slow convergence speed;
Using the genetic algorithm's global search capacity, the network enhances the connection weights and thresholds of the BP network so that it searches from a better initial value and finds the global optimum more rapidly;
The model can make a more accurate prediction of the unknown graduation grades by using the existing students' course grades;
Apply the optimized model to the student achievement management system, and use the prediction results to give academic warnings to students.

The following issues remain, nonetheless, with this study:

Gathering of representative data. The student course achievement sample size used in this study is insufficient. The accuracy of the prediction model can be improved by boosting the sample size. Second, in order to increase the scope of research elements in performance prediction, more pertinent components can be added since the indicators of sample data are insufficient;

The model's enhancement. In this study, a genetic algorithm is used to optimize the BP neural network model's pre-parameters. The post parameters can be adjusted using a different approach since they are learned using the traditional least squares algorithm but are not optimized. We shall keep getting better in the upcoming research in light of the inadequacies mentioned above.

Funding

This paper was not funded by any project or organizations.

Competing Interests

All authors declare that there is no conflict of interest regarding the publication of this paper.

Acknowledgement

The authors would like to thank the anonymous reviewers who have provided valuable comments on this article.

Author Contributions

Xia Liu contributed to the study conception and design. Jing Guo contributed to Material preparation, data collection and analysis were performed by Jingjing Jiang. The first draft of the manuscript was written by Xia Liu and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Data Availability

Data were obtained upon a reasonable request.

Ethical approval

This article does not contain any studies with human participants or animals performed by any of the authors.

Witten, Frank IH, Data mining (2005) Practical Mach Learn Tools Techniques Java Implementations 13(1):1–1
Witten IH, Frank E (2011) Data mining: practical machine learning tools and techniques. Acm Sigmod Record 31(1):76–77
Daniel Sanchez JM, Serrano I, Blanco (2012) Maria Jose Martin-Bautista Maria-Amparo dependencies. Vila. Using association rules to mine for strong approximate Data Mining and Knowledge Discovery. 8(3):321 ~ 326
Romero C, Ventura S, E García (2008) Data mining in course management systems: Moodle case study and tutorial. Comput Educ 51(1):368–384
Romero C, Ventura S (2007) Educational data mining: A survey from 1995 to 2005[J]. Expert Syst Appl 33(1):135–146
Ayesha S, Mustafa T, Sattar AR et al (2010) Data mining model for higher education system,European Journal of entific Research, (1)
Tissera W, Athauda RI, Fernando HC (2007) Discovery of Strongly Related Subjects in the Undergraduate Syllabi using Data Mining, International Conference on Information & Automation. IEEE,
Sun H, Research on Student Learning Result System based on Data Mining (2010) Int J Comput Sci Netw Secur 10(4):203–205
Xianhong L (2012) Application of WPF to MIS development, 2012 International Conference on Information Management, Innovation Management and Industrial Engineering. IEEE, 2: 223–226
Gorla N, Somers TM, Wong B (2010) Organizational impact of system quality, information quality, and service quality. J Strateg Inf Syst 19(3):207–228
4th International Workshop on Wearable and Implantable Body Sensor Networks (2007) : March 26–28, 2007 RWTH Aachen University, Germany, Springer Science & Business Media, 2007
Fei G, Gao J, Owodunni O et al (2011) A method for engineering design change analysis using system modelling and knowledge management techniques. Int J Comput Integr Manuf 24(6):535–551
Mingers J, White L (2010) A review of the recent contribution of systems thinking to operational research and management science. Eur J Oper Res 207(3):1147–1161
Robles K, Fraga A, Morato J et al (2012) Towards an ontology-based retrieval of UML Class Diagrams. ‎Inf Softw Technol 54(1):72–86
Khare A, Lam H (2008) Assessing student achievement and progress with online examinations: Some pedagogical and technical issues. Int J E-learning 7(3):383–402
Long Y (2019) Research on art innovation teaching platform based on data mining algorithm. Cluster Comput 22(6):14943–14949
Romero C, Ventura S (2013) Data mining in education, Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, 3(1): 12–27
Algarni A (2016) Data mining in education,International Journal of Advanced Computer Science and Applications, 7(6)
Kirimi JM, Moturi CA (2016) Application of data mining classification in employee performance prediction. Int J Comput Appl 146(7):28–35
Ahmed TM, Bezemer CP, Chen TH et al (2016) Studying the effectiveness of application performance management (apm) tools for detecting performance regressions for web applications: an experience report, 2016 IEEE/ACM 13th Working Conference on Mining Software Repositories (MSR). IEEE, : 1–12
Qin J (2017) Design and implementation of sports performance management system in higher vocational colleges based on data mining, The 1st EAI International Conference on Multimedia Technology and Enhanced Learning.
Zhang R, Tao J (2017) A nonlinear fuzzy neural network modeling approach using an improved genetic algorithm. IEEE Trans Industr Electron 65(7):5882–5892

Download PDF

Journal Publication

published 04 Apr, 2023

Read the published version in Soft Computing →

Editorial decision: Major Revision
01 Feb, 2023
Reviewers agreed at journal
15 Jan, 2023
Reviewers invited by journal
15 Jan, 2023
Editor assigned by journal
11 Jan, 2023
First submitted to journal
10 Jan, 2023

You are reading this latest preprint version

The Design of College Student Achievement Management System based on GA-BP Network

Status:

Journal Publication

Version 1

Abstract

Figures

1 Introduction

2 Related Works

3 Performance Prediction Model Based On Ga-bp Neural Network

3.1 The establishment of BP neural network model

3.2 Establishment of GA-BP Neural Network Model

3.2.1 Choice of encoding

3.2.2 Design of fitness function

3.2.3 Genetic manipulation

3.2.4 Stop condition

3.3 GA-BP network hybrid algorithm flow

3.4 Design of student achievement management platform based on GA-BP network model

4 Result Analysis And Discussion

4.1 Dimension reduction of sample data

4.2 Application of GA-BP neural networks in students' performance prediction

5 Conclusions

Declarations

References

Status:

Journal Publication

Version 1