Data-driven Civic Education: Theory and Practice

In the current situation, data has become a fundamental strategic resource in today’s world, and the incorporation of data resources into the construction of ideology and politics will be of great benefit to ideology and politics to meet the requirements of the new era and promote its own development.

Big data continues to promote the digitization of social life, but its data security, data privacy, data alienation and other difficult problems of the times arise impacting the theoretical and practical boundaries of ideological and political education, and the balance between instrumentality and value of data-driven in the era of big data requires us to make a positive response [1]. Admittedly, the development of things has two sides, we can fully explore the positive aspects of the development of the advantages of maximization. The problem of data is not only the problem of data itself, but also the problem of “human” thought and behavior, the combination of data-driven and ideological and political education is conducive to play the advantages of data, to solve the problem of data from the source, and to seek a new situation for the development of ideological and political education [2-3]. At present, the innovative consciousness of ideological and political education is still insufficient, the utilization of modern science and technology still needs to be strengthened, the concept of data-driven has not yet been emphasized, and most of them carry out their work in the traditional way, neglecting the powerful function of combining with the big data and ignoring the possibilities of innovative development [4-6]. The important quality of advancing with the times is the driving force for things to keep moving forward, and the continuous and steady development of Civic Education should also advance with the times. Therefore, facing big data in the new era, seizing the historical opportunity brought by data-driven, utilizing data to drive the development of ideological and political education in depth, utilizing technology to enhance precision in data-driven, focusing on value guidance in data-driven, and enhancing the temperature and efficacy of data [7-9]. First of all, it is necessary to have an in-depth understanding of the systematic theory of data-driven. It is a necessary way to improve the theoretical system of ideological and political education in colleges and universities, and is conducive to promoting the development of the discipline of ideological and political education. Data-driven ideological and political education will deeply analyze the foundational theories that examine the nature of ideological and political education, interactive relationship, content structure, effective path, emotional tendency and so on [10-11]. Secondly, it promotes cross-disciplinary mutual reference and integration, and enhances the scientific and inclusive nature of ideological and political education in colleges and universities. The data-driven research understanding undoubtedly enhances the scientificity of the discipline, accommodates the relevance data with an inclusive mindset, and depicts the sample of things, which provides a brand-new perspective for the research of the discipline of ideology and politics [12-14]. Finally, data-driven in methodological perspective enriches and enriches the theoretical study of ideological and political education methods [15]. Big data has become a hot issue in academic research, and more and more scholars apply big data as a method to solve the real problems of ideological and political education, which provides conditions for the proposal and application of data-driven methods.

This paper explores the application of data-driven technology in civic education through the theoretical analysis of the integration of data-driven and civic education. Selecting agricultural and water majors as the case object, the K-prototype clustering algorithm is used to cluster and analyze the big data of students’ civic education based on the data collected from the questionnaire survey, and construct student portraits based on the analysis results. Based on the learning characteristics and learning needs of students with different profiles derived from clustering, data mining technology is used to recommend personalized civic education resources suitable for students with different needs from the civic education resource database. At the same time, the similarity between students in the recommendation model is calculated to prevent anomalies caused by student similarity in the recommendation process of educational resources. The hierarchical analysis method is used to assign weights to the indicators in the teaching effect assessment index system of Civic and Political Education for Agricultural and Water Professions, and each teaching indicator is divided into quantitative grades. The least squares support vector machine algorithm is used to construct a model for evaluating the teaching effect of Civics and Political Education in Agricultural and Water Sciences, which is used to evaluate the teaching effect of Civics and Political Education in this specialty in a scientific and comprehensive way.

2

Overview

Literature [16] examines how realistic data-driven approaches can improve college students’ physical fitness index and conduct integrated teaching of physical education.The essence of education concepts is to cultivate students’ correct three views and knowledge systems, so this result can also be utilized in civic education. Literature [17] mentions that the data-driven utilization of online multimedia and digital virtual technology provides innovations in Civic and Political Education, which has a slight impact on students’ political development and personal thoughts. And literature [18] mentions a data-driven program that facilitates the initiation of actions by teachers to identify and develop quality character for specific situations. The results of this study provide a good exercise for teachers of political education to continuously maintain their best qualities for better political education.

In addition, the literature [19] utilized a data-driven approach to the degree of open-mindedness of the relevant population during Trump’s presidency, and the results showed that these populations are characterized by a statistically significant distribution of open-mindedness, which does not waver over time. It shows that changing one’s mind is not easy, therefore better education on ideology is needed to help people form a correct ideological system.Similar to literature [20], a data-driven investigation was conducted to examine the relationship between perceived collective threats and political preferences. The research showed that the political left and right differ in their approach to threats.Based on this result, it can be used in civic education to categorize students with different ideological and political preferences, rationalize educational preferences, and ensure the correct functioning of the civic system. In addition, literature [21] introduces to the Twitter social platform that the political right publishes more false information and extreme statements than the left. And literature [22] also mentions that symbolic, connectionist, and data-driven artificial intelligence programs are not neutral or objective.It can be seen that data-driven information lacks objectivity and impartiality, and is prone to inducing people’s ideological politics. Therefore, literature [23] evaluates the political and ethical aspects of value creation in data-driven educational practices, which can make students aware of algorithmic control so that they can treat data-driven more correctly and rationally.

3

Data-driven and civic education

3.1

Applications of big data in education

The core concept of data-driven education theory is the use of educational data to support decision making and curriculum improvement [24]. This theory is based on theories from the fields of statistics, data analysis methods, and machine learning. By collecting data on students’ academic performance, learning behaviors, and classroom participation, educational institutions can gain a better understanding of students’ learning trends and difficulties. The theory of data-driven education suggests that the analysis of this data can guide educators in adjusting course content, teaching methods, and resource allocation to improve student academic achievement.

The theory of personalized learning emphasizes that each student is unique and that the educational experience should be designed according to their academic level, interests, and learning pace.In the age of big data, personalized learning is becoming more feasible by analyzing large-scale student data. Schools and edtech companies can use this data to create personalized learning paths, recommend learning resources, and provide personalized feedback. This personalized approach encourages students to be more actively engaged in their learning and improves their academic performance.

3.2

Integration of Big Data and Civic Education

The theoretical convergence of big data and Civics education builds on the theoretical foundations of data-driven education and Civics education. This convergence emphasizes the use of big data technologies to enhance the content and methods of education in civics and political science classes.The theoretical foundation focuses on gathering and analyzing data on students’ political awareness, values, and social participation to gain a better understanding of their ideological and political development.For example, by analyzing data on students’ participation in social activities, educators can better understand their students’ social responsibility and civic literacy.

Although the theoretical integration of big data and education in ideology and politics courses provides new opportunities for education, it is also accompanied by a number of problems and challenges from a theoretical perspective. These aspects include balancing privacy protection and ethical principles for educational data, and ensuring that data-driven education emphasizes not just academic performance but also students’ ideological and political development. At the same time, educators need to consider how to integrate big data technologies in Civics and Politics courses to improve the quality of education while ensuring that the moral and ethical principles of education are adhered to.

4

Data-driven applications in Civic Education in Agricultural Water Programs

As a discipline with strong application, the Civic and Political Education of Agricultural Water Conservancy Specialty should be closely integrated with the characteristics of the specialty, focusing on cultivating students’ sense of social responsibility and practical ability. Based on the theoretical analysis in the previous section, this section will take the agricultural water specialty as an example to explore the application method of data-driven in the ideological and political education of agricultural water specialty.

4.1

Data collection and analysis

4.1.1

Data collection

A questionnaire survey was carried out using the whole cluster sampling method among some agricultural and water majors in five vocational colleges and universities, namely, Xuzhou Institute of Industrial Vocational Technology, Jiangsu Institute of Architecture Vocational and Technical College, Jiangsu Institute of Safety Technology Vocational College, Xuzhou Early Childhood Normal Higher and Specialized School, and Xuzhou Institute of Technicians of Jiangsu Province. A total of 4673 questionnaires were returned, 23 unqualified questionnaires were excluded, and the validity rate of the questionnaires was 99.51%. Among them, 2374 were male students and 2276 were female students. By grade, there were 1643 first-year students, 1539 second-year students and 1468 third-year students. The survey was conducted using a self-administered questionnaire, which was released on the Questionnaire Star online survey platform and sent to students to fill in through a QR code.The content of the questionnaire mainly includes basic student information such as gender and age, historical learning data of the Civic Education Program, and ideological and political quality.

4.1.2

Data analysis

In this paper, we analyze the big data of agricultural and water students’ civic education by clustering based on K-prototype algorithm, and construct students’ portrait on the basis of clustering [25]. K-prototype algorithm inherits the ideas of K-means algorithm and K-modes algorithm, and adds the prototype of describing the data clusters and the formula for calculating the degree of dissimilarity between the data of mixed attributes, and it is a typical algorithm for dealing with mixed attribute clustering typical algorithm.

1)

Distance Measure

K-prototype measures numerical and categorical features according to K-means and K-modes distance calculations Euclidean distance and Hamming distance, respectively, which are combined to form the distance to the prototype. Assuming that the sample is n and the set of m features D = {X₁, X₂, ⋯, X_n}, X_i and X_j represent two samples respectively.

For numerical type features, it is first necessary to normalize them and map them to [0,1] intervals, and then calculate the distance of numerical features, K-means uses Euclidean distance derived from the formula for the distance between two points in the Euclidean space, and the distance is expressed as: (1) $d_{1} (X_{i}, X_{j}) = \sum_{i = 1}^{m_{r}} {(x_{i t}^{r} - x_{j t}^{r})}^{2}$

For category type features, K-modes are computed using the Hamming distance: (2) $d_{2} (X_{i}, X_{j}) = \sum_{i = 1}^{m_{l}} δ (x_{i t}^{t}, x_{j t}^{t})$ where when p = q, δ(p,q) = 0. When p ≠ q, δ(p,q) = 1. For sample i, $x_{i t}^{r}$ and $x_{i t}^{r}$ are the numerical features, $x_{j t}^{t}$ and $x_{j t}^{t}$ are the category features, and m_r and m_t are the number of numerical and category features, respectively.

Calculating the dissimilarity between objects of mixed feature types can be done by combining different features into a single dissimilarity matrix, Let k be the number of clusters and Q_t = {q_s1,q_s2,⋯,q_sm} denote the category center chosen for category c, so the distance between the data and the center cluster can be expressed as follows: (3) $d (X_{i}, Q_{j}) = d_{1} (X_{i}, Q_{j}) + γ_{i} d_{2} (X_{i}, Q_{j})$

Then the loss function of the K-prototype can be defined as: (4) $L = \sum_{c = 1}^{k} (L_{s}^{r} + L_{s}^{r}) = L^{r} + L^{r}$

$L_{s}^{r}$ then represents the total loss of all numerical features in the sample of category c , $L_{s}^{r}$ represents the total loss of all category features, and γ_s is the weight of the category features in category c . γ_s affects the accuracy of clustering. When γ_s = 0, only numerical features are considered, which is equivalent to the K-means method, and when γ_s is larger, category features account for more weights, and the clustering results are dominated by categorical variables. Therefore, choosing a suitable γ_s can lead to better clustering. The choice of γ_s is affected by the mean square deviation of the numerical variables, and when the mean square deviation is set to 1, γ_s it is better to set it to 0.3~0.7.

2)

Hybrid clustering K-prototype

The numerical variables are standardized to have variance 1, so γ_c is set to 0.5. The specific steps of the K-prototype algorithm are as follows:

Step 1: Randomly select k initial clustering centers {c₁,c₂,⋯,c_k} from dataset D.

Step 2: Traverse the dataset D, calculate the distance between the sample and each clustering center according to Equation (3), and assign the sample to the category closest to the center.

Step 3: Update the clustering centers of numerical features and category features after each sample allocation. Formula (1) is used to calculate for numerical features and formula (2) is used for category features.

Step 4: The loss function is calculated using equations (3) and (4).

Step 5: If the new loss function value is less than the set threshold or the number of obtained bands is greater than the set T, then the calculation is finished and the clustering center is output, otherwise repeat steps 2, 3 and 4.

3)

Cluster characterization and portrait construction

The selection of the number of clusters is related to the different clusters to get the statistical characteristics and portrait, and the profile coefficient combines the degree of cohesion and separation, which can objectively evaluate the clustering effect. Therefore, by setting a different number of clusters, the clusters with the best performance in contour coefficient are selected.

After forming the clusters, the center of each cluster is described, and for the numerical features, the mean and variance of the student group of the cluster are calculated, and the evaluation is carried out in three dimensions, including basic attributes, and the clusters are further subdivided into fine-grained divisions according to the average grades of the courses of the Civic and Political Education, so as to make the construction of the portrait of these clusters more comprehensive and specific.

4.2

Recommendations for Civic and Political Education Resources

Based on the results of data analysis, teachers can develop personalized teaching content for different students.For example, students with lower ideological and political quality may benefit from increased relevant course content and practical activities.For students who already have higher ideological and political qualities, they can be guided to participate in higher levels of social practice and volunteer activities. In this section, data mining algorithms will be used to recommend suitable learning resources for students to meet the learning needs of students at different levels for the content of ideological and political education.

4.2.1

Modeling

The introduction of Civics and Politics teaching resources promotes the innovative development of agricultural water education, promotes the cross-fertilization of agricultural water education and humanities and social sciences, and provides strong support for the cultivation of agricultural water talents with innovative spirit and interdisciplinary background. Therefore, the agricultural water courses are personalized to recommend the Civic and Political teaching resources. First of all, in order to ensure that the recommendation effect of the Civics teaching resources is targeted, the Civics teaching resources recommendation model for agricultural water courses is now constructed to analyze the learning behavior and the corresponding Civics teaching resources.

After obtaining the historical learning data of agricultural and water majors, the recommendation model of Civic and political education resources for agricultural and water major courses is expressed in terms of and binary group with the following formula: (5) $k = (h_{i j}^{n}, l_{a s}^{m})$ Where: h_ij is the trajectory of the use of the Civic Teaching Resources of the Civic Education Course in Agricultural and Water Sciences, $l_{a s}^{m}$ is the threshold value of the course learning behaviors, ij is the position of each teaching resource in the model, n is the number of the trajectory of the use of the Civic Teaching Resources, m is the number of the learning behaviors in the model, and as is the percentage of a certain kind of learning behaviors in the model, at this time, the formula of the directed graph of the use of the Civic Resources of the computer-based course trajectory $h_{i j}^{n}$ is as follows: (6) $h_{i j}^{n} = (q, w)$ Where: q is the node of ideological and political teaching resources in the agricultural and water courses in the model, and w is the edge between the nodes of teaching resources. Considering the dynamic process of students’ course learning, after completing the behavioral classification, the weights of various learning behaviors are decomposed and calculated to provide a basis for the allocation and recommendation of teaching resources in the future.

4.2.2

Feature extraction

Set the keywords in the trajectory of the use of the Civic and Political Teaching Resources in the nst Agricultural and Water Program as v. A keyword set is constructed based on the historical search keywords of the students in the model, and data mining is used to update the learner’s search keyword weights, which are given in the following formulas: (7) $\partial = B^{\frac{ln (T - t)}{r}} + Y_{v} \times α \times β$ Where: ∂ is the weight of the keywords of the learning interest node of the Civic and Political Teaching Resource in the Agricultural and Water Program, and $B^{\frac{h (T - t)}{r}}$ represents the evaluation function. r is the network evaluation factor of the teaching resource, T is the current date of the model of the recommended resource, t is the date of the most recent modification of the keywords in the recommendation model, Y_v is the collection of keywords for historical learning, β is the learning rate, and α is the corresponding score of the evaluation of the teaching resource.

The similarity between students in the recommendation model is calculated to avoid anomalous recommendation results caused by similarity of students in the future recommendation process of teaching resources. Therefore, the similarity formula between students M and N is given as follows: (8) $Z (f_{M}, f_{N}) = \frac{\sum_{i = 1}^{5} Y_{v}^{M} * Y_{v}^{N} * g_{i}^{2}}{\sqrt{\sum_{i = 1}^{5} {(Y_{v}^{M} * g_{i})}^{2}} \sqrt{\sum_{i = 1}^{5} {(Y_{v}^{N} * g_{i})}^{2}}}$ Where: $Y_{v}^{M}$ is the set of historical keywords for student M, $Y_{v}^{N}$ is the set of historical keywords for student N, and g_i is the weight of student feature information.

4.2.3

Personalized Recommendations

Define a collection of teaching resources that is designed to meet the needs of students, denoted as c = {c_i,i ∈ 1,2,…,n}, where n denotes the total number of candidate teaching resources available for selection. For each candidate teaching resource to be selected c_i, two key metrics need to be evaluated: one is the frequency of the resource being used by students u. The other is the average recommendation result that the resource receives from the recommender system η. Combining the above information, the recommendation index of each teaching resource is calculated μ , and the candidate teaching resources are ranked according to the recommendation ratio, and the best teaching resources are recommended to the students according to the following formula: (9) $p (c_{i}) = ϖ u (c_{i}) + (1 - ϖ) μ (c_{i})$ Where: ϖ is a variable parameter in the calculation process.

In practical application, the situation that the recommended indexes of the two alternative teaching resources are exactly the same will be encountered, therefore, the two alternative teaching resources must be judged twice, and the formula is as follows: (10) $X = \sum_{i = 1}^{n} η (s_{i}, p (c_{i}))$ Where: s_i is the recommendation vector of teaching resources.

The above formula is used to make secondary determination of educational resources in order to achieve accurate recommendation of ideological and political educational resources in agricultural and water courses, so as to provide personalized learning resources for students.

4.3

Evaluation of teaching effectiveness

4.3.1

Teaching effectiveness assessment system

In order to objectively assess the teaching effect of Civic and Political Education in Agricultural Water Specialties, the index system for assessing the teaching effect of Civic and Political Education in Agricultural Water Specialties is firstly established, based on the principles of science, objectivity, practicability, easy data collection, etc., the index system for evaluating the teaching effect of Civic and Political Education in Agricultural Water Specialties is established as shown in Table 1.

Table 1.

Evaluation system for effectiveness of ideological and political education

Target layer	Primary indicator	Secondary indicator	Number
Evaluation indicator system for the effectiveness of ideological and political education in agricultural and water major	Ideological and political quality (X₁)	Political attitude	x₁
		Values	x₂
		Social responsibility	x₃
	Academic performance (X₂)	Normal grade	x₄
	Academic performance (X₂)	Final grade	x₅
	Practice activity participation (X₃)	Social practice number	x₆
	Practice activity participation (X₃)	Volunteer service duration	x₇
	Teaching content and method (X₄)	Satisfaction with course content	x₈
	Teaching content and method (X₄)	Satisfaction with teaching method	x₉
	Teacher quality and ability (X₅)	Teacher’s expertise	x₁₀
	Teacher quality and ability (X₅)	Teacher teaching experience	x₁₁
	School environment (X₆)	Campus culture construction	x₁₂
	School environment (X₆)	Academic atmosphere	x₁₃
	Student feedback and satisfaction (X₇)	Students’ satisfaction with teaching	x₁₄
	Student feedback and satisfaction (X₇)	Students’ satisfaction with course	x₁₅
	Employment situation (X₈)	Graduate employment	x₁₆
	Employment situation (X₈)	Employment quality of graduates	x₁₇

4.3.2

Teaching Effectiveness Assessment Leveling

The assessment of the teaching effect of Civic and political education in agricultural and water majors is a quantitative treatment of the teaching effect, which needs to be quantified for each teaching index in order to be modeled, and this paper adopts a 100-point system to classify the teaching effect of Civic and political education into five grades, and the specific correspondences are shown in Table 2.

Table 2.

Classification of teaching effects

Grade number	Grade name	Corresponding score
1	Excellence	90~100
2	Good	80~90
3	Medium	70~80
4	Qualify	60~70
5	Out of line	<60

4.3.3

Determination of indicator weights

1)

Establishment of judgment matrix

Through the comparison between the upper and lower levels to determine the degree of importance of different indicators, this paper adopts 1-9 values to scale the degree of importance of the indicators for evaluating the teaching effectiveness of agricultural and water education in the field of ideology and politics.

2)

Hierarchical single sorting

Each column of the judgment matrix is normalized and calculated to obtain the weight of the evaluation indicators of the teaching effectiveness of Civic and Political Education, and the weight calculation formula is specified as follows: (11) $ω_{i} = \frac{1}{n} \sum_{j = 1}^{n} (a_{i j} / \sum_{k = 1}^{n} a_{i j})$ Where: n indicates the number of columns, i.e., the number of indicators for evaluating the teaching effectiveness of Civic and Political Education in Agricultural and Hydraulic Programs. a_ij indicates the importance of indicator i relative to indicator j.

3)

Judge the consistency of the matrix

In practical application, the consistency of the matrix needs to be tested, and the judgment matrix is reasonable only if it satisfies the consistency. The consistency formula is: (12) $C I = \frac{λ_{max} - n}{n - 1}$ Where λ_max denotes the maximum weight value.

4)

Consistency test

By checking the table to get the order of judgment matrix, calculate the average stochastic consistency index (RI), and calculate the consistency ratio (CR) by Eq. (13), if its value is less than 0.1, then it means that the judgment matrix is reasonable, and it also means that the obtained weights are valid: (13) $C R = C I / R I$

Determine the weight value of the indicators for the assessment of the teaching effectiveness of agricultural and water education in the field of ideology and politics through the hierarchical analysis method, and the results are shown in Figure 1, in which the size of the sphere indicates the weight of the indicators. From the figure, it can be seen that the consistency of each indicator is less than 0.1, indicating that the weights of the indicators calculated are valid. In the evaluation index system of the teaching effect of Civic and Political Education in Agricultural Water Specialties, the highest weight of the index of final grade (x₅) is 0.109, and the lowest weight of the index of volunteer service hours (x₇) is 0.012.

4.3.4

Least Squares Support Vector Machines

To address the problem of low learning efficiency of support vector machine, this paper proposes the least squares support vector machine [26], which changes the constraints of the support vector machine as follows: (14) $y^{k} [ω^{\cdot} φ (x_{k}) + b] = 1 - e_{k}$

A transformation of Eq. (14) yields the equivalent quadratic optimization form as follows: (15) ${min}_{α, k} \frac{1}{2} ω^{T} ω + \frac{γ}{2} \sum_{k = 1}^{N} e_{k}^{2}$

Lemma Lagrange multiplier α_k, which establishes the Lagrange function, i.e., there: (16) $\begin{array}{l} L (ω, b, e, α) = \frac{1}{2} ω^{T} ω \\ + \frac{γ}{2} \sum_{k = 1}^{N} e_{k}^{2} - \sum_{k = 1}^{N} α_{k} {y_{k} [ω^{\cdot} φ (x_{k}) + b] - 1 + e_{k}} \end{array}$

According to Eqs. (17) to (20), the values of Lagrange multiplier α_k and bias vector b can be obtained.

(17)

\frac{\partial L}{\partial ω} = 0 \to ω = \sum_{k = 1}^{N} α_{k} y_{k} φ (x_{k})

(18)

\frac{\partial L}{\partial b} = 0 \to \sum_{k = 1}^{N} α_{k} y_{k} = 0

(19)

\frac{\partial L}{\partial e_{k}} = 0 \to α_{k} = e α_{k}

(20)

\frac{\partial L}{\partial α_{k}} = 0 \to y_{k} [ω^{\cdot} φ (x_{k}) + b] - 1 + e_{k} = 0

Introducing the kernel function instead of the inner product operation: K(x,x_i) = φ^T(x)φ(x), then the decision function of the least squares support vector machine is: (21) $f (x) = \sum_{i = 1}^{N} α_{i} K (x, x_{i}) + b$

4.3.5

Ideas for assessing teaching effectiveness

The idea of evaluating the teaching effect of Civic and political education of agricultural water specialty based on data mining is as follows: firstly, establish the index system for evaluating the teaching effect of Civic and political education of agricultural water specialty, and use hierarchical analysis method to determine the weight of each index. Then collect the relevant data of the indicators of the evaluation of the teaching effect of agricultural water major’s civic and political education, and determine the teaching effect level, and establish the learning sample of the evaluation of the teaching effect of agricultural water major’s civic and political education. Finally, the least squares support vector machine is used to rate the teaching effect of the Civic and political education of agricultural water majors, and the teaching effect evaluation model of the Civic and political education of agricultural water majors is established.

5

Practical applications and analysis

In the previous section, this paper proposed the application methods of data-driven in the Civic and Political Education of Agricultural and Water Professions, which mainly include the use of clustering algorithm to construct student portrait, data mining algorithm to personalize and recommend the Civic and Political Education resources, and evaluation of the teaching effect of Civic and Political Education based on the Hierarchical Analysis Method and the Least-Squares Support Vector Machine algorithm, and so on. In this section, we will analyze the application of these methods to the practice of Civic and Political Education in agricultural and water majors through examples, verify the effectiveness of the methods proposed in this paper, and highlight the role of data-driven methods in Civic and Political Education.

5.1

Student Portrait Construction

5.1.1

Data processing

In this paper, the data of agricultural and water students were collected through a questionnaire, in order to make the data In order to make the data fit the algorithm better, the study used modern information processing techniques, and in the pre-processing process, invalid features and more missing data items were further removed. Table 3 shows the data and attributes used in this study including 4650 students and 13 attributes in the dataset. In addition, Figure 2 shows the correlation between the attributes. From the figure, it is easy to conclude that the grade point average (y₇ - y₁₃) for each semester maintains a high correlation with ideological and political qualities, while other attributes maintain a low correlation with ideological and political qualities.

Table 3.

Dataset description

Categories	Describe	Symbol
Basic data	Gender	y₁
	Peoples	y₂
	Political appearance	y₃
Honor	Get school honor	y₄
	Receive above provincial honors	y₅
	Total amount of money awarded	y₆
Grade	First semester	y₇
	Second semester	y₈
	Third semester	y₉
	Fourth semester	y₁₀
	Fifth semester	y₁₁
	Sixth semester	y₁₂
	Total score	y₁₃

5.1.2

Parameter analysis and model comparison

1)

Experimental setup

This study uses the K-prototype algorithm to build multiple clusters of transformed dataset attributes as needed for use. For this purpose, the following settings are used in this study:

(1) Operating system: Microsoft Windows 10 (64-bit operating system, X-64 based processor).

(2) Python: Python 3.9 was used to transform the data from its original state to a new dataset that could be fed into the algorithm, complete with algorithm implementation, statistical analysis and imaging.

2)

Model comparison and optimal K-value

The number of clusters is a predetermined parameter when applying the K-Prototype algorithm. The goal of clustering is to cluster data of the same category in the dataset together as much as possible and separate data of different categories as much as possible. Therefore, before implementing clustering, the K-value needs to be optimized. For this problem, in the study, the numerical features were first normalized to the interval [0,1], and then the numerical features were mixed with the categorical features as input data for input. Finally, the K was set to a positive integer between [1,10], and the maximum number of iterations of the algorithm was specified as 500 for each K value.

Figure 3 shows the clustering profile coefficients when global optimization is achieved. The K-Prototype algorithm is analyzed by comparing it with K-means algorithm and DBSCAN in order to confirm that the method proposed in this paper is most suitable for the research scenario. The results of the study show that the K-Prototype algorithm outperforms the K-means algorithm at all K settings and the DBSCAN algorithm at K = 3. Specifically, K-means uses only numerical data, DBSCAN uses the optimal values at different clustering radii, and the K-Prototype algorithm has been tested for γ = 0.4. Based on the results shown in the figure, it is clear that the best contour coefficient is obtained when K = 3 and the best value is 0.471. For K-means algorithm, there is a decreasing trend in the clustering performance which may be due to the data dimensionality. For DBSCAN, on the other hand, after changing its clustering radius, the algorithm did not show a significant tuning trend and its performance remained around 0.4, which is lower than the optimal performance of the K-Prototype algorithm. Therefore, the study obtained the optimal K value of 3 for implementing the K-Prototype algorithm, which means that three student groups are sufficient to cluster all students in the study.

5.1.3

Model interpretation

From the results of the above figure, it can be seen that when K = 3, the profile coefficient of the cluster is 0.471, which is the best among all the clusters, indicating that K = 3 is the optimal number of clusters for the object of study in this paper. Therefore, the number of clusters of the K-Prototype algorithm in the study was set to 3. Finally, three groups of students were obtained, labeled as Group 1 (G1), Group 2 (G2), and Group 3 (G3), with 1,673, 2,182, and 795 students in each group, respectively. The group characteristics of the three groups of students obtained will be specifically analyzed below.

Table 4 shows the results of descriptive statistics and significant differences in the characteristics of the student population, where ***, **, and * denote 1%, 5%, and 10% significance levels, respectively. The normality of each feature was first tested and the results showed that all features did not obey the normal distribution, so the Kruskal-Wallis test could be performed. The results of the test showed that most of the features in the clustering results showed significant differences. Among them, the difference in ideological and political qualities features (y₇ - y₁₃) is the most obvious, all of them showing different significance levels of 1%. This is followed by the difference in the performance of honors and awards (y₄ - y₆), which has the next highest level of difference. In contrast, the difference in basic student characteristics (y₁ - y₃) was the least significant. Overall, the results of the study indicate that the differences between the three student groups are significant, suggesting that the research methodology is applicable to the scenarios in this study.

Table 4.

Results od descriptive statistics and significance tests

	G1	G2	G3	Kruskal-Walls test
y₁	0.53	0.47	0.22	0.009**
y₂	0.27	0.38	0.31	0.083*
y₃	0.58	0.22	0.33	0.103
y₄	0.64	0.57	0.26	0.005**
y₅	0.19	0.28	0.19	0.002**
y₆	0.23	0.31	0.27	0.003**
y₇	0.77	0.64	0.58	0.000***
y₈	0.83	0.68	0.61	0.000***
y₉	0.79	0.73	0.59	0.000***
y₁₀	0.91	0.66	0.63	0.000***
y₁₁	0.85	0.79	0.54	0.000***
y₁₂	0.93	0.75	0.60	0.000***
y₁₃	0.89	0.71	0.62	0.000***

According to the clustering results, further analysis of the characteristics of each group found that the ideological and political quality of the students in group G1 is significantly better than that of the students in groups G2 and G3, and the ideological and political quality of the students in group G2 is better than that of the students in group G3. Therefore, these three groups of students can roughly be divided into three groups: high ideological and political quality, medium ideological and political quality, and low ideological and political quality. For the convenience of the subsequent expression, in this paper, we uniformly describe the students of the G1 group as the excellent ideological and political quality group, the students of the G2 group as the medium ideological and political quality group, and the students of the G3 group as the poor ideological and political quality group.

5.2

Analysis of Recommended Resources for Civic and Political Education

5.2.1

Data preparation

In order to verify that the method of this paper can effectively personalize the recommendation of Civics and Political Education resources, the method of comparative testing is used to complete the demonstration. Select hybrid recommendation method and deep learning-based recommendation method as the control group respectively, take the students who have completed the clustering and grouping in the above section as the test objects, and select thousands of groups of ideological and political education resources from the resource library as the test data to verify the recommendation effect of different methods. In order to ensure the accuracy of the data test, the Civic and political education resources are retrieved from the resource library, and thousands of groups of data are randomly selected as recommendation samples according to the ideological and political literacy of different groups, as shown in Table 5.

Table 5.

Education resource data sample

Resources	G1	G2	G3
A1	1037	112	103
A2	1075	118	92
A3	225	1154	111
A4	210	1093	96
A5	104	83	1201
A6	100	110	1212
A7	109	103	86
A8	1100	100	119
A9	213	1211	1225
A10	212	91	95

5.2.2

Recommended accuracy of different methods

Here, different recommendation methods are used to recommend Civic Education resources so that they can be taught in different groups for Civic Education.The MATLAB test platform is used to upload the selected resource samples and connect them to three groups of recommendation methods.The results of the recommendation accuracy test for each group of methods are shown in Fig. 4. As can be seen from the figure, the recommendation accuracy rate of this paper’s method on the three groups is 0.94, 0.93 and 0.88 respectively, which is higher than the other two methods. The hybrid recommendation method has the second highest accuracy rate, and the deep learning method has the lowest recommendation accuracy rate for the Civic Education resources in the research scenario of this paper. It shows that the method in this paper can recommend the Civic and Political education resources that meet the learning characteristics of different groups of agricultural and water students according to their ideological and political literacy and learning needs.

5.2.3

Resource Liking Test under Personalized Recommendation

In order to further analyze the effect of the role of different recommendation methods, with the selected students as the test object, set the length of a single course of Civic and Political Education Resources under each type as 45 min, respectively, to verify the degree of fondness of different groups of students for the recommended resources, and the results are shown in Table 6. As can be seen from the table, in this paper’s method of recommending different types of Civic and Political Education resources, all of them can be maintained at more than 42 min listening time, which indicates that college students in each group have a high degree of favoritism towards the recommended resources and can accept this type of Civic and Political Education. At the same time, when using this paper’s method for recommending civic education resources, the length of listening time of G3 students with lower ideological and political quality is higher than that of the other two groups, which indicates that the civic education resources recommended by this paper’s method are more in line with the learning needs of the G3 group of students, and can play an important role in improving the ideological and political quality of the G3 group of students. The listening time under the other two methods is less than 40 min, and the listening time under the deep learning method is generally less than 20 min, which shows that the method in this paper can not only meet the good matching of the educational resources of Civics and Politics, but also ensure the students’ enjoyment of Civics and Politics courses, and has a better recommending effect, which can be put into the actual Civics and Politics teaching and applying.

Table 6.

Lecture duration statistics (min)

Group	Resources	Ours	Mixed recommendation	Depth learning
G1	A1	43.48	22.37	17.26
	A2	42.76	22.05	17.58
	A8	43.03	21.36	18.31
G2	A3	42.87	22.44	15.63
	A4	42.51	21.18	16.22
	A9	43.01	21.39	16.08
G3	A5	44.56	17.65	9.84
	A6	43.98	13.56	9.17
	A7	44.21	11.34	9.26
	A10	44.35	16.49	9.08

5.3

Evaluation of Civics Teaching Effectiveness

5.3.1

Parameter setting

The mean square error is selected as the evaluation index, and the output results of the teaching effect assessment model used in this paper’s model under different Lagrange multipliers are counted, and the statistical results are shown in Fig. 5. From the experimental results, it can be seen that when the Lagrange multiplier is 0.8, the output result of this paper’s model has the lowest mean square error, indicating that the model has a high output performance at this time, and when applied to the assessment of the teaching effectiveness of the Civics and Political Science of Agricultural and Hydraulic majors, it can effectively improve the assessment accuracy, and the Lagrange multiplier of setting up the Least Squares Support Vector Machine Model applied to the Civics and Political Science teaching assessment is 0.8.

5.3.2

Assessment of results

The hierarchical analysis method is used to calculate the grade classification of the assessment indicators of the teaching effect of Civic and Political Education in Agricultural Water Specialties, and the results are shown in Figure 6. As can be seen from the figure, the grade of each teaching effect assessment index is “qualified” and above, of which the grade of final grade (x₅) and teacher professionalism (x₁₀) index is classified as “excellent”.Social responsibility (x₃), number of social practices (x₆), length of volunteer service (x₇), teachers’ teaching experience (x₁₁), and academic research atmosphere (x₁₃) are graded as “good”.The overall rating for the assessment of the effectiveness of the teaching of Civic Education in the Agricultural Water Program is “Good”.The results show that the model in this paper can effectively realize the assessment of the teaching effect of Civic and political education in agricultural and water professions, and provide reference for the improvement of the teaching effect of Civic and political education in agricultural and water professions.

The results of evaluating the teaching effect of Civics in agricultural water majors using least squares support vector machine are shown in Table 7. The experimental results can be seen, using the model of this paper can realize the evaluation of Civics teaching effect, and the final evaluation result is 84 points, indicating that the Civics teaching effect of agricultural and water majors in the selected institutions of this paper is good, and there is a certain amount of room for improvement. Institutions should put forward relative improvement measures for the evaluation results from the indicators with lower scores such as teaching content and methods (X₄) and employment situation (X₈) to improve the teaching effect of Civics and Political Education in Agricultural Water Specialties.

Table 7.

Evaluation results of our model

Primary indicator	Scoring	Secondary indicator	Scoring
X₁	78	x₁	72
		x₂	67
		x₃	83
X₂	83	x₄	71
X₂	83	x₅	94
X₃	88	x₆	82
X₃	88	x₇	86
X₄	69	x₈	68
X₄	69	x₉	73
X₅	91	x₁₀	93
X₅	91	x₁₁	87
X₆	73	x₁₂	65
X₆	73	x₁₃	82
X₇	71	x₁₄	77
X₇	71	x₁₅	73
X₈	66	x₁₆	64
X₈	66	x₁₇	72
Total	84

6

Conclusion

This paper explores the direction and methods of the application of data-driven technology in Civic Education through the specific application of K-prototype clustering algorithm, data mining technology, least squares support vector machine and other methods in the Civic Education of agricultural and water majors.

1) The results of descriptive statistics and significant differences in the characteristics of the student groups show that when the optimal number of clusters K = 3 was selected, the differences in the characteristics of the students in terms of ideological and political qualities were significant, and all the indicator variables were significant at the level of 1%. And further analysis of the differences in the characteristics of students’ ideological and political qualities shows that the ideological and political qualities of students in the G1 group are significantly better than those of students in the G2 and G3 groups, with scores ranging from 0.77-0.93. The ideological and political qualities of students in the G2 group are the next best, and the ideological and political qualities of students in the G3 group are the lowest in comparison. Therefore, according to the results obtained by the model, the students majoring in agriculture and fisheries selected in this paper were divided into three groups, and named them as “high ideological and political quality”, “medium ideological and political quality” and “low ideological and political quality” according to the differences in ideological and political quality.

2) The recommendation accuracy rate of this paper’s Civics education resources recommendation method on the three groups of G1, G2 and G3 is 0.94, 0.93 and 0.88 respectively, which is higher than that of the hybrid recommendation method and the Civics education resources recommendation method based on deep learning. At the same time, the recommended Civics education resources of this paper’s method fully satisfy the Civics learning needs and preferences of different groups of students, and the three groups of students can maintain the length of listening time at 42min. And the listening time of the G3 group is longer than that of the G1 and G2 groups. It shows that the recommended model of Civics education resources in this paper can play an important role in enhancing the interest of Civics learning of agricultural and water science majors and improving the performance of Civics courses of students.

3) In the grading of each index of the teaching effect evaluation system of agricultural and water professional ideological and political education by analytic hierarchy process, the grades of the two indicators of “final grades” and “teacher professionalism” were “excellent”, the grades of five indicators such as “social responsibility” were “good”, and the overall grade was “good”. On the whole, the scores of “teaching content and methods” and “employment situation” in the evaluation of the teaching effect of ideological and political education in agriculture and water majors are 69 and 66 respectively. Colleges and universities can improve the effect of ideological and political education in agriculture and water majors by strengthening the construction of projects related to these two indicators, and use data-driven technology to promote the development of ideological and political education in agriculture and water majors.

Fund Project:

Zhejiang University of Water Resources and Electric Power’s 2023 University-Level Key Course Project “Descriptive Geometry and Engineering Drawing”.

Język:: Angielski

Częstotliwość wydawania:: 1 razy w roku
Dziedziny czasopisma:: Nauki biologiczne, Nauki biologiczne, inne, Matematyka, Matematyka stosowana, Matematyka ogólna, Fizyka, Fizyka, inne

Kanał RSS czasopisma

Data-driven Civic Education: Theory and Practice

Qian Wang

Yonggang Duan

Data publikacji: 21 mar 2025

Otrzymano: 18 lis 2024

Przyjęty: 20 lut 2025

DOI: https://doi.org/10.2478/amns-2025-0666

Słowa kluczoweData-driven, K-prototype, Data mining, Least squares support vector machine, Civic education

© 2025 Qian Wang et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Słowa kluczowe
Data-driven, K-prototype, Data mining, Least squares support vector machine, Civic education