^{1}

^{2}

^{3}

^{4}

^{1}

^{2}

^{3}

^{4}

Sustainable development education respects differences and encourages different assessment methods to evaluate students. During the epidemic, many colleges’ examinations changed from offline to online. How to fully consider students’ process learning status and make a reasonable evaluation of students is worthy of research. Based on the process learning data of a course in a university in China, this study establishes a discrete Hopfield neural network model to classify the test samples. In the process of modelling, the grey correlation analysis method is used to optimize the elements affecting students’ comprehensive evaluation index, and it solves the problem of failure of the model due to the large gap between the factors in the traditional discrete Hopfield neural network model. Then, the entropy right TOPSIS method is used to rank samples with the same evaluation grade. Teachers can objectively evaluate each student’s process learning performance according to the ranking results. Finally, the article compares and analyzes the evaluation results of various different methods. The analysis results believe that the optimized discrete Hopfield neural network is feasible in the process learning evaluation, and the model evaluation results are more objective and comprehensive.

Education plays an important role in human beings. The United Nations General Assembly has advocated that governments should take sustainable development education as an education strategy and action plan. In recent years, many countries have incorporated the concept of sustainable development education into curricula, textbooks, and educational teaching practices and have been fruitful. Traditional curriculum learning evaluation tends to pay more attention to the result evaluation, ignoring the process evaluation and individual differences. Affected by the epidemic, China’s education quickly moved from offline to online, so that teachers’ “teaching” and students’ “learning” have undergone unprecedented changes. Teachers and students mainly adopt online teaching and examination through the teaching platform, so how to integrate process learning evaluation and final examination evaluation and make an objective and fair evaluation of students is very worthy of study.

Process learning evaluation, also known as formative evaluation, is an immediate, dynamic, and repeated evaluation of students in the teaching process, which focuses on timely feedback to strengthen and improve students’ learning [

Hopfield neural network is a recurrent neural network, proposed by J. Hopfield in 1982. The discrete Hopfield neural network in Hopfield neural network is named DHNN (discrete Hopfield neural network). It is a single-layer feedback neural network with feedback connection from output to input; it is also a kind of cyclic neural network [

In discrete Hopfield neural networks, some optimization algorithms are often combined with discrete Hopfield neural networks in order to make them more capable of association with memory. The improved method has a heuristic effect on other complex recurrent neural networks. Many scholars have changed the network structure, improved the weight design algorithm [

Affected by the epidemic, in the first half of 2020, Chinese universities mainly use online teaching and examination: how to integrate process evaluation and final examination evaluation organically and give students fair, objective, and reasonable evaluation is necessary. In this context, if teachers still take students’ final examination results as the evaluation standard of students’ learning this course, it is clear that the evaluation results are not objective enough. The development process from quantity change to qualitative change is long, and learning evaluation is a dynamic process. Therefore, this research takes the process assessment of college students’ curriculum learning as the research content, through the analysis of students’ dynamic learning process state, using qualitative and quantitative analysis methods, and it establishes an effective evaluation model and gives an objective and fair evaluation.

This study took 49 students in the second year of a university as the research object, and the research data came from the learning process data of the “Education theory” course in the background of mobile interactive teaching software (xue-xi-tong APP) (time: the second semester of the 2019–2020 academic year). For the first time, 6 indicators were selected, and the initial classification level was given by the comprehensive evaluation score of learning software, as shown in Table

The dynamic learning process ability level and the corresponding evaluation index of 49 students.

Number | Watching audio and video | Times of chapter learning | Score for homework | Score for unit test | Score for group task | Course interaction | Grade (ABCDE) |
---|---|---|---|---|---|---|---|

1 | 100.0 | 100.0 | 95.0 | 73.5 | 92.7 | 90.0 | A |

2 | 95.4 | 100.0 | 94.0 | 84.1 | 93.3 | 100.0 | A |

3 | 100.0 | 100.0 | 98.2 | 75.8 | 90.0 | 100.0 | A |

4 | 71.6 | 100.0 | 97.2 | 88.6 | 90.0 | 100.0 | A |

5 | 96.2 | 89.0 | 95.9 | 71.2 | 90.0 | 89.0 | B |

6 | 74.0 | 90.0 | 92.6 | 73.5 | 92.7 | 90.0 | B |

7 | 96.0 | 92.0 | 95.6 | 71.2 | 73.3 | 94.0 | B |

8 | 71.6 | 91.0 | 99.0 | 71.2 | 71.7 | 88.0 | B |

9 | 74.5 | 92.0 | 96.9 | 84.1 | 73.3 | 90.0 | B |

10 | 70.8 | 42.4 | 86.3 | 84.1 | 86.7 | 91.0 | B |

11 | 74.0 | 92.0 | 91.0 | 79.5 | 73.3 | 92.5 | B |

12 | 73.0 | 97.6 | 91.7 | 65.2 | 92.7 | 78.0 | B |

13 | 72.0 | 86.6 | 96.6 | 71.2 | 71.7 | 92.5 | B |

14 | 73.4 | 74.2 | 82.7 | 77.3 | 81.7 | 65.0 | C |

15 | 74.0 | 100.0 | 92.6 | 84.1 | 61.0 | 89.0 | C |

16 | 70.8 | 43.4 | 95.4 | 18.2 | 90.0 | 100.0 | C |

17 | 72.0 | 38.4 | 89.8 | 71.2 | 85.0 | 82.5 | C |

18 | 76.0 | 47.6 | 91.0 | 76.5 | 87.7 | 92.5 | C |

19 | 71.0 | 100.0 | 91.8 | 84.1 | 61.0 | 90.0 | C |

20 | 73.4 | 68.8 | 83.1 | 75.8 | 90.0 | 72.5 | C |

21 | 71.0 | 53.4 | 80.8 | 79.5 | 72.7 | 75.0 | C |

22 | 70.8 | 35.0 | 89.9 | 84.1 | 73.3 | 80.0 | C |

23 | 75.0 | 68.8 | 88.8 | 84.1 | 86.7 | 77.5 | C |

24 | 71.0 | 100.0 | 91.3 | 79.5 | 61.0 | 100.0 | C |

25 | 72.0 | 50.0 | 91.3 | 84.1 | 72.7 | 82.5 | C |

26 | 72.0 | 71.0 | 97.0 | 79.5 | 71.7 | 62.5 | C |

27 | 72.0 | 100.0 | 85.8 | 79.5 | 86.3 | 62.5 | C |

28 | 72.0 | 91.6 | 95.7 | 68.9 | 80.0 | 57.5 | C |

29 | 70.0 | 95.0 | 88.7 | 33.3 | 73.3 | 95.0 | C |

30 | 70.8 | 100.0 | 94.8 | 71.2 | 56.7 | 90.0 | C |

31 | 73.4 | 70.8 | 93.8 | 67.4 | 71.7 | 77.5 | C |

32 | 73.0 | 38.4 | 92.1 | 75.8 | 71.7 | 82.0 | C |

33 | 70.8 | 26.6 | 86.0 | 84.1 | 91.7 | 52.0 | C |

34 | 72.3 | 100.0 | 82.7 | 75.8 | 73.3 | 67.5 | C |

35 | 74.2 | 34.2 | 85.6 | 84.1 | 73.3 | 82.0 | C |

36 | 90.0 | 89.2 | 92.9 | 93.2 | 77.7 | 40.0 | C |

37 | 73.0 | 100.0 | 84.5 | 79.5 | 66.7 | 60.0 | C |

38 | 72.0 | 59.2 | 95.0 | 61.4 | 70.3 | 72.5 | C |

39 | 74.2 | 65.0 | 87.9 | 81.1 | 76.0 | 65.0 | C |

40 | 72.0 | 29.2 | 81.0 | 77.3 | 78.3 | 85.5 | C |

41 | 70.8 | 60.0 | 70.0 | 84.1 | 73.3 | 58.0 | C |

42 | 72.0 | 25.8 | 85.2 | 33.3 | 73.3 | 95.0 | C |

43 | 76.8 | 80.8 | 86.3 | 68.9 | 70.3 | 65.0 | C |

44 | 74.2 | 92.4 | 73.3 | 71.2 | 65.0 | 35.0 | D |

45 | 70.8 | 48.4 | 84.3 | 56.1 | 63.3 | 67.5 | D |

46 | 73.4 | 83.4 | 86.6 | 71.2 | 57.7 | 55.0 | D |

47 | 70.4 | 75.0 | 74.7 | 75.8 | 76.7 | 57.5 | D |

48 | 70.8 | 15.0 | 75.7 | 71.2 | 74.3 | 57.5 | D |

49 | 70.0 | 17.6 | 36.0 | 41.7 | 63.3 | 22.5 | E |

Suppose the output value of DHNN is −1 or 1, which are recorded as the inhibition and excitation states of neurons, respectively. The given marks are as follows:

The initial structure of the DHNN is composed of six neurons, as shown in Figure

Structure diagram of discrete Hopfield network.

Set the working mode of the Hopfield network to serial mode. It is assumed that the working network is stable. The rule of evolution is to decrease the energy function until it reaches a stable state. The Lyapunov function here is the energy function, and it is defined as follows:

In this model, the outer-product method is used to design the Hopfield network, and the training goal preserves K

The operation steps are as follows:

Step 1: initialize the network.

Step 2: the ith neuron is randomly selected from the network.

Step 3: calculate the input value

Step 4: calculate the output value

Step 5: to determine whether the network is stable or not: if it is stable or meets the given conditions, it ends; otherwise, go to step 2 and continue.

The steady state here is defined as

The ideal grade evaluation index is designed.

The average of the evaluation indicators corresponding to the samples in Table

The ideal grade evaluation index is coded.

As shown in the first line of Figure

The test sample index is coded.

Select ten samples from Table

MATLAB is used to create the discrete Hopfield neural network.

The simulation results are in Figure

As seen from Figure

Five-grade ideal evaluation indexes.

Grade | Index | |||||
---|---|---|---|---|---|---|

A | 91.8 | 100 | 96.1 | 80.5 | 91.5 | 97.5 |

B | 78 | 85.8 | 95.4 | 74.6 | 80.6 | 89.1 |

C | 73.1 | 65.9 | 88.5 | 72.8 | 75.3 | 77.7 |

D | 71.9 | 62.8 | 82.9 | 68.6 | 69.4 | 54.5 |

E | 70 | 17.6 | 36 | 41.7 | 63.3 | 22.5 |

Evaluation index values of 10 test samples.

Original serial number | New serial number | Grade | ||||||
---|---|---|---|---|---|---|---|---|

1 | Example 1 | 100.0 | 100.0 | 95.0 | 73.5 | 92.7 | 90.0 | A |

6 | Example 2 | 74.0 | 90.0 | 92.6 | 73.5 | 92.7 | 90.0 | B |

8 | Example 3 | 71.6 | 91.0 | 99.0 | 71.2 | 71.7 | 88.0 | B |

10 | Example 4 | 70.8 | 42.4 | 86.3 | 84.1 | 86.7 | 91.0 | B |

12 | Example 5 | 73.0 | 97.6 | 91.7 | 65.2 | 92.7 | 78.0 | B |

36 | Example 6 | 90.0 | 89.2 | 92.9 | 93.2 | 77.7 | 40.0 | C |

43 | Example 7 | 72.0 | 25.8 | 85.2 | 33.3 | 73.3 | 95.0 | C |

44 | Example 8 | 74.2 | 92.4 | 73.3 | 71.2 | 65.0 | 35.0 | D |

48 | Example 9 | 70.8 | 15.0 | 75.7 | 71.2 | 74.3 | 57.5 | D |

17 | Example 10 | 72.0 | 38.4 | 89.8 | 71.2 | 85.0 | 82.5 | C |

Simulation results of grade evaluation for 10 test samples.

The basic idea of grey relational analysis is to judge whether the relation is close or not according to the geometric shape similarity of the sequence curve. The closer the curves are, the greater the correlation between the corresponding sequences is. The steps of grey correlation analysis are as follows:

Step 1: determine the system analysis sequence.

The reference sequence is denoted as

Comparison sequence (also known as subsequence) is an effective factor affecting the main behavior of the system. There are six effective factors here. The behavior sequence of factor

Step 2: the dimensionless treatment of variables.

Because of the different scale of the data affecting the various factors of the comprehensive evaluation, in order to facilitate comparison, it is necessary to carry out the nonscale processing

Step 3: calculate the comprehensive correlation degree [

Grey system modelling software can also be used here (software downloads address:

Step 4: build a new evaluation index.

Evaluating Indicator

Evaluating Indicator

Evaluating Indicator

Put it all together and fill in Table

Step 5: construct a new ideal classification index hierarchy.

According to the value of each case in Table

Evaluation index values of 10 test samples.

New serial number | |||
---|---|---|---|

Example 1 | 100 | 87.1 | 90 |

Example 2 | 83.6 | 86.3 | 90 |

Example 3 | 83.2 | 80.6 | 88 |

Example 4 | 53.8 | 85.4 | 91 |

Example 5 | 87.8 | 79 | 78 |

Example 6 | 89.5 | 87.9 | 40 |

Example 7 | 44.3 | 63.9 | 95 |

Example 8 | 85.1 | 69.8 | 35 |

Example 9 | 37.3 | 75 | 57.5 |

Example 10 | 51.8 | 82 | 82.5 |

Adjusted ideal grade evaluation index.

Grade | Index | ||
---|---|---|---|

A | 96.7 | 89.4 | 97.5 |

B | 82.7 | 83.5 | 89.1 |

C | 69.6 | 78.8 | 77.7 |

D | 67.2 | 74.1 | 54.5 |

E | 38.6 | 47 | 22.5 |

Repeating the steps in 2.2.3, the ideal grade evaluation index and waiting classification index is coded again. Running MATLAB code, the simulation results are organized as shown in Figure

Classification simulation results of the reconstructed discrete Hopfield neural network.

It can be seen from Figure

As can be seen from Figure

According to the formula of entropy weight TOPSIS, the final result is obtained:

The results are summarized in Table

Comparative analysis table of evaluation results of 10 students to be classified.

Original serial number | Initial classification level | Grading by optimized neural network | Score of entropy weight TOPSIS Ci | Ranking of entropy weight TOPSIS | Final exam results | Grading by final exam |
---|---|---|---|---|---|---|

1 | A | B | 98.17 | 1 | 86.0 | B |

6 | B | B | 86.76 | 2 | 90.0 | A |

8 | B | B | 78.66 | 3 | 94.0 | A |

10 | B | B | 67.6 | 4 | 88.0 | B |

36 | B | B | 60.66 | 5 | 66.0 | D |

12 | C | C | 73.03 | 1 | 66.0 | D |

17 | C | C | 56.72 | 2 | 96.0 | A |

48 | D | D | 23.44 | 1 | 50.0 | E |

43 | C | E | 38.28 | 1 | 50.0 | E |

44 | D | E | 32.91 | 2 | 60.0 | D |

It can be seen from Table

Grade A: no student has reached this level.

Grade B:

In the three different evaluation methods, the evaluation level of the number 10 student is consistent. It shows that the student is at a good level no matter the kind of evaluation method. Because the student usually has good study habits, the examination results are also ideal.

Number 1 student usually studies very seriously; especially, their self-discipline ability is better. But this students’ application ability and classroom interaction ability is a little poor, which is consistent with his final grade.

Although the test score of number 36 student belongs to D level, the self-discipline ability and practical ability of this student are relatively good, so it is classified as B level. At the same time, we found that he had less interaction with the teacher in class, indicating that the student was shy and not good at expressing himself, so he should pay more attention to the classmate in the future.

Grade C:

Although the final result of student number 17 was very good, his attitude towards learning was slow. He had a low score for watching videos and completing task points, and he did not cooperate with the teacher in the learning process. In future, the teacher should deeply understand the needs of students, timely adjust the way of class, and let all students participate in the course learning. On the other hand, the student’s final exam results are questionable, and we may need to know more about the student’s learning of other subjects and other students’ evaluation of him.

Grade D:

Grade E:

In addition, the entropy weight TOPSIS method gives the same level of student ranking, which provides a strong reference for students’ usual scores. For example, there are five students in class B. According to the entropy weight TOPSIS score column in Table

Through the above analysis, the following conclusions are drawn:

Conclusion 1: reasonable process learning evaluation method can objectively evaluate students’ normal performance justly, dynamically grasp students’ normal learning state, analyze students’ learning characteristics from students’ learning state, and provide valuable help to students targeted.

Conclusion 2: the optimized discrete Hopfield neural network model is effective in the process learning evaluation. Through the grey correlation analysis among various factors, the weighted processing of impact evaluation indexes according to the correlation degree can achieve the goal of model optimization. In particular, it solves the problem that the traditional discrete Hopfield neural network model fails to find the equilibrium point due to the large gap between factors.

Conclusion 3: the method of entropy weight TOPSIS can objectively give the order of the samples with the same level, which provides a good reference for the given results of teachers in peacetime.

Education is a major force for sustainable development and change and for improving people’s ability to transform social ideals into reality. Education for sustainable development has been seen as a process of learning how to make decisions that take into account economic, ecological, and all social equity. Cultivating this future-oriented thinking ability is a key task of education. This paper has established that sustainable development education has built up students’ dominant position in the teaching process. Especially in the online teaching background, students’ self-discipline, practical ability, and classroom interaction ability have been fully reflected. The dynamic process learning evaluation model combined with the three evaluation indexes gives the evaluation results of students’ daily learning objectively and comprehensively. This result is helpful to do a good job of education and teaching and to provide reference and guidance for students’ education management.

The data used to support the findings of this study are available from the authors upon request.

The authors declare that they have no conflicts of interest.

C. S., Y. M., and T. C. were responsible for conceptualization; Y. M. prepared methodology; Y. M. curated data; C. S., Y. M., and T. C. wrote the original draft; C. S. and Y. M. carried out review and editing; Y. M., C. S., and T. C. performed visualization; and C. S. and T. C supervised the data.

The authors acknowledge the Research Project of Humanities and Social Sciences in Colleges and Universities of Anhui Province (SK2019A0545); Online Important Teaching Research Project of Anhui Province Colleges and Universities (2020zdxsjg234); Anhui Quality Engineering Project (2018mooc508 and 2019mooc273); and Teaching Team Project of Chaohu University (ch20jxtd02).