Research on the Combination Strategy of Marxist Ideology and Ideological and Political Education in Colleges and Universities Based on Data Fusion Modeling

In recent years, data fusion technology has been increasingly widely used in several fields, especially in the field of education, where it can provide new ideas and methods for educational reform with its powerful data processing and analyzing capabilities. By integrating and analyzing information from different sources, data fusion technology can help us understand students’ learning status, ideological dynamics, and changes in the social environment more comprehensively and accurately. This provides a practical program for deepening the combination of Marxist thought and ideological and political education in colleges and universities [1].

At present, the research on ideological and political education mainly focuses on educational concepts, teaching methods, and course content. Scholar Zhu Hong pointed out in Research and Application of Campus Data Platform Based on Data Integration that the effectiveness of ideological and political education not only relies on the teaching level of teachers, but also is closely related to the student’s sense of identity, participation, and social environment [2]. Lai Ling in “The Innovative Integration of Student Management and Ideological and Political Education in Colleges and Universities in the Era of Big Data” explains that the traditional education model often ignores the individual differences of students and the influence of social factors, resulting in unsatisfactory educational effects. In the field of data fusion, relevant research is gradually increasing. Scholar Li Yingzhen mentioned in “Using Big Data Technology to Realize Intelligent Ideological and Political Education in Colleges and Universities” that data fusion can effectively improve the scientificity and accuracy of educational decision-making. Yan Dongming mentioned in “Digital Twin Dynamic Model Extrapolation Technology Based on Multi-source Heterogeneous Data” that data fusion technology is used to analyze students’ learning behaviors to optimize teaching strategies. However, the research for ideological and political education is still insufficient, especially on how to apply the data fusion model specifically to the teaching practice of Marxist ideology, which needs to be explored in depth [3].

In this paper, we study the combination strategy of Marxist ideology and ideological and political education in colleges and universities based on the data fusion model and construct the data fusion model through the weighted average algorithm, Kalman filter algorithm, and decision tree algorithm, to identify the key factors affecting the effect of ideological and political education. The aim is to propose targeted educational strategies, which provide a new theoretical basis for ideological and political education, as well as practical programs for educational practice in colleges and universities.

2

Data Fusion Model Algorithm

2.1

Weighted average method

The weighted average method [4] is a simple and commonly used method for static data fusion. It integrates the eigenvalues of multiple data sources by assigning weights to different data sources. It is suitable for processing datasets where qualitative indicators (e.g., satisfaction, agreement, etc. as shown in Figure 1) are more explicit.

There are ⁿ data sources ^{(x₁,x₂.....,x_n)}, The corresponding weights are ^{(W₁,w₂,...,w_n)} The formula (1) for the weighted average method is: (1) $F = \frac{\sum_{i = 1}^{n} w_{i} x_{i}}{\sum_{i = 1}^{n} w_{i}}$

^x_i : Observations from the first fear, ⁱ data source.

^wⁱ: The important weight of the i data source.

The weighted average algorithm is used to fuse quantitative [5] survey data (e.g., student satisfaction and engagement). For example, student engagement came from three sources: classroom interactions (weighted 0.4), online course recordings (weighted 0.35), and questionnaires (weighted 0.25). Classroom interaction scores: 4.2, 4.4, 4.1 Online course scores: 3.8, 4.0, 3.9 Questionnaire scores: 4.5, 4.6, 4.4.

Substitution formula(2): (2) $F = \frac{0.4 \cdot 4.2 + 0.35 \cdot 3.8 + 0.25 \cdot 4.5}{0.4 + 0.35 + 0.25} = 4.06$ The results showed a fused score of 4.06, indicating a high level of overall student engagement.

2.2

Kalman filter

Kalman filtering is a recursive algorithm [6] that is widely used for the estimation of dynamic data. It optimizes the effect of noise and obtains more accurate results by weighted averaging the current observations (e.g., Figure 2) with the predicted values. It is suitable for time series data such as multiple questionnaires or classroom performance over time. The recursive formula (3) for Kalman filtering is as follows: (3) $\hat{x} k = \hat{x} k - 1 + K_{k} (z_{k} - H {\hat{x}}_{k - 1})$

{\hat{x}}_{k}

: Estimated value of the state at the current moment k

{\hat{x}}_{k - 1}

: The state estimate of k_k–1 the previous moment Z_i : The observed value of ^k at the current moment K_k : Kalman gain, calculated as: (4)

K_{k} = \frac{P_{k - 1} H^{T}}{H P_{k - 1} H^{T} + R}

Where ^P(k−1) is the prediction error covariance matrix, H is the observation matrix, and ^Ris the observation noise covariance matrix? The algorithm is used to dynamically monitor changes in student engagement. For example, the observed values of students’ weekly classroom engagement are as follows::

\begin{matrix} z_{1} = 3.8, & z_{2} = 4.0, & z_{3} = 4.3 \end{matrix}

Prediction of the initial value of^{$({\hat{x}}_{0} = 3.7)$} , through the Kalman filter for dynamic correction of the degree of participation, the calculation results are:

{\hat{x}}_{1} = 3.78,, {\hat{x}}_{2} = 3.96,, {\hat{x}}_{3} = 4.12

It can be seen that the degree of student participation is a gradual increase in the trend, indicating that the adjustment of teaching strategies has been effective.

2.3

Decision tree

Decision tree algorithms [7] learn the relationship between inputs and outputs by constructing rules based on data features. Compared to traditional methods, machine learning models can handle nonlinear data and complex feature interactions. The decision tree formula is shown below:

The classification process of the decision tree can be described as maximizing the information gain (5) $I G (D, A) = H (D) - \sum_{v \in Values (A)} \frac{| D_{v} |}{| D |} H (D_{v})$ (6) $H (D) = - \sum_{i = 1}^{n} p_{i} \log_{2} p_{i}$

The entropy of the data set ^D ^D: : The subset of data corresponding to a given value ^ν of attribute ^A ^P_i : Probability of class ⁱ in the data set.

It is used to fuse multiple features (e.g., student cognition, affective tendencies, etc.) to predict overall educational outcomes. For example, feature inputs include: ^x₁ : Student satisfaction (quantitative value) ^x₂ : Course attendance (percentage) ^x₃ : Participation in online discussions (qualitatively coded values) By training the decision tree model, it is possible to predict the increase in overall student satisfaction after course optimization. When applied to the data, the actual predicted increase in satisfaction was 25%, and the accuracy of the model was 92% as shown in Figure 3.

2.4

Algorithm Comparison and Optimization

The results based on data testing are as follows: Table 1.

Comparison of algorithms

Algorithm	Advantages	Limitations	Accuracy (%)
Weighted average	Simple and easy to implement	Cannot handle dynamic or complex relationships	85
Kalman filtering	Performs well with dynamic data	Requires accurate noise modeling	90
Decision tree method	Can handle complex, multidimensional features	Sensitive to data volume	92

The comparison allows you to choose the appropriate fusion method according to the data characteristics:

Static data: preferred weighted average method.

Dynamic data: use Kalman filtering.

Data with complex features: use machine learning algorithms.

3

Ideological and political education based on data fusion model

3.1

Data layer fusion

Data layer fusion is the basis of the data fusion model, and its main task is to collect and integrate effective information from multiple data sources [8]. Ideological and political education in colleges and universities needs to comprehensively consider students’ classroom performance, online behavior, extracurricular activities participation, and other factors as shown in Figureure 4, so the data sources mainly include the following three categories:

Classroom data: Collect students’ attendance records, classroom participation (e.g. speaking frequency, degree of interaction), learning outcomes, etc. through the smart classroom system.

Online learning data: Based on the data of online learning platforms (e.g. MOOC platform, on-campus course management system), students’ learning progress, homework submission, forum interaction content, etc. are acquired.

Social behavior data: supplement the deep information of students’ ideological dynamics through questionnaires, records of students’ club activities, and psychological counseling files.

At this stage, the data undergoes ETL technology [9] to build the data warehouse, the data within the data warehouse is mainly formed by extracting the data from each single data source loaded into the data warehouse, and ETL can realize this extraction function. The specific process of this calculation is [10]: firstly, extract data from each single data source (sub-business system), and then convert the extracted data, and then carry out data cleansing to eliminate all kinds of errors in the data, such as missing values, duplicates, etc., to facilitate the realization of the data regularity; finally, the processed data can be refreshed to the target database. ETL technology is the key technology of data fusion, which belongs to the previous and next technology. When building a shared data platform, the first step is to understand what the school has and what type of business it has, and then consider how to extract data from these single data sources; after that, the data is cleaned; finally, it is designed and integrated, and then converted into a unified analysis format. The main task of data layer fusion is to extract uniformly structured information from multiple sources of data to provide a basis for subsequent analysis. Figure 4 shows the ETL technology roadmap.

Heterogeneous datasets with multiple sources are available: ^{D¹,D₂,D₃,...D_n} Indicates raw data from different data sources. The features are represented by a matrix: each data source ^D_i consists of ^m_i records and ^n_i features, represented as a matrix ^{X_i ∈ Rm_i×n_i} Data preprocessing Data Cleaning: for each dataset X_i, remove missing values, and noisy data, and fill in missing items as follows: (7) $X_{i}^{'} = Clean (X_{i})$ Data normalization: to avoid differences in feature magnitude of different data sources affecting the fusion results, all features are normalized as in Eq (8): (8) $X_{i}^{n o r m} = \frac{X_{i}^{'} - μ_{i}}{σ_{i}}$ μ_i and are the mean and standard deviation ^{$X_{i}^{'}$} , respectively.

Heterogeneous Data Integration Assume that ^{D¹,D²,D³,…D_n} has ^{P₁,P₂,…,P_n} common features respectively, which are integrated into a unity matrix as in Eq(9): (9) $X = \oplus_{i = 1}^{n} X_{i}^{norm}$ For example, if the common characteristics of classroom behavior data and online learning data are engagement time ^T and interaction frequency ^F , then as Eq: (10) $X = [T_{1}, F_{1} \dots T_{2}, F_{2} \dots]$

3.2

Feature layer fusion

The core task of feature layer fusion [11] is to extract key features reflecting students’ state of mind from multidimensional data (as shown in Figure. 6) Through the cross-analysis and feature extraction of data from multiple sources, the model can identify students’ thought dynamics and their behavioral tendencies. Specific methods include [12]:

Sentiment analysis: natural language processing of textual content in students’ online learning interactions (e.g., course discussion forum speeches, social platform sharing) to identify their emotional tendencies and value expressions. Calculate the effective tendency score 1 from Eq: S: 11 $S = \sum_{k = 1}^{n} w_{k} □ f_{k} (w_{k}) (f_{k})$

Behavioral pattern analysis: using machine learning algorithms (support vector machines and random forests) to model classroom participation data and online behavioral data to analyze students’ ideological and behavioral patterns, such as whether they have a positive attitude toward participation or are at risk of deviating from the socialist core values, as shown in Eq. 12 $y = sign (w^{T} x + b)$ ^w is the weight vector, ^b is the bias, and ^y indicates the category of thought dynamics (e.g, positive/neutral/negative)

Classification of student groups: students are classified with different ideological characteristics through clustering algorithms, which provides a basis for the subsequent design of personalized education.

Through feature layer fusion, the data were further processed into core variables with explanatory power, such as the level of student’s knowledge of core Marxist concepts and their acceptance of core socialist values[13].

3.3

Integration at the decision-making level

Decision-level fusion is the final output stage of the data fusion model [14], whose main goal is to support ideological and political education decision-making through comprehensive analysis. Machine learning models are utilized to predict the trend of students’ ideological changes using the decision tree algorithm: (13) $I G (D, A) = H (D) - \sum_{v \in Values (A)} \frac{| D_{v} |}{| D |} H (D_{v})$ Through optimization calculation, the optimal resource allocation scheme can be generated for different thought characteristic groups. Differentiated teaching strategies are designed for students of different ideological characteristic groups [15]. For students who are active but easily influenced by external network culture, the basic theoretical teaching of Marxist ideology is strengthened; while for students who are cognitively mature but lack the power of action, more emphasis is placed on the design of practical sessions. The model predicts the risk points of students’ ideological fluctuations and provides school administrators with intervention strategies. For example, when certain student groups are found to have deviated from their values, thematic education activities are carried out promptly. The model is used to analyze the effects of educational measures, such as curriculum reforms and practical activities, on the changes in students’ thinking, thus providing a basis for the continuous improvement of educational policies.

4

Empirical research and case studies

4.1

Data sources and study design

To verify the application effect of the data fusion model in ideological and political education in colleges and universities, this study takes a key university as a case study to carry out empirical research. The study collects data from multiple sources, constructs a fusion model, analyzes the ideological dynamics of students, and optimizes educational policies based on the analysis results [15]. The specific design includes the following links:

1)

Research object

In this study, 1,000 undergraduate students of the class of 2024 in a university were selected as the research subjects with the aim of exploring the differences in ideological and political education among students of different majors and grades and their impact on learning outcomes. The participants all volunteered to participate in this study and filled out an informed consent form prior to their participation to ensure the ethicality of the study and the participants’ right to know. There were 500 participants in each of the experimental and control groups, and the sample was designed to cover a wide range of disciplines, including arts, science and technology, and arts, to ensure diversity and representativeness of the study population. The 10-week data collection cycle will contribute to an in-depth understanding of how students change and develop in the process of ideological and political education. The basic characteristics of the study population show that the participants come from diverse backgrounds, with some of the students having been educated in Marxist theory at the university level, while others have only received a basic political course. This diverse educational background will enable us to analyze the impact of different forms of ideological and political education on students’ perceptions and attitudes, thus providing an empirical basis for the optimization of ideological and political education in universities. Through systematic analysis of the research data, we expect to reveal the differences in the needs of students of different majors in ideological and political education, and provide guidance for colleges and universities in curriculum design and teaching strategies.

2)

Data sources and collection methods

The study combines quantitative data with qualitative data collection, and the data sources include the following three categories: Classroom behavior data: students’ classroom participation is recorded through the smart classroom system, including attendance, speaking frequency, and discussion activity. Online learning platform data: extracting students’ course learning trajectories, homework completion, online discussion records, and other data from the online learning platform of a university. Questionnaire and interview data: A standardized questionnaire was designed for students’ ideological orientation, including students’ understanding of Marxist theory; their identification with socialist core values; students’ satisfaction with the ideological and political courses, and suggestions for improvement. Meanwhile, 30 students were randomly selected for in-depth interviews to supplement the meticulousness of the data. The types of data include the following three dimensions: Classroom interaction data (CI): the number of questions asked, the discussion participation rate, and the correct rate of classroom answers.

Online learning data (OL): viewing hours, completion rate, test scores.

Questionnaire feedback data (QF): satisfaction scores (1-5).

Table 2.

Table of data sources

Data type	Characterization	Date Range	Data type	Unit
CI	Number of questions	0-10	Continuous	Times
	Discussion Participation Rate	0%-100%	Percentage	%
	Correct Answer Rate	0%-100%	Percentage	%
OL	Watching time	0-20	Continuous	h
	Course Completion Rate	0%-100%	Percentage	%
	Test Score	0-100	Continuous	Point
QF	Satisfaction Rating	1-5	Discrete	mark

3)

Data analysis methods

In this study, Python and SPSS were used as data analysis tools [16], and the following methods were mainly adopted: feature extraction: students’ online discussion records were analyzed by natural language processing technology to extract emotional tendencies (positive/neutral/negative) and thought characteristics; Cluster analysis: using K-means [17]algorithm such as formula to classify students with different ideological characteristics; (14) $\min_{μ_{1} μ_{2} \dots μ_{k}} \sum_{i = 1}^{m} \sum_{K = 1}^{K} r_{i, k} □ {| x_{i} - μ_{k} |}^{2}$ ^{x_i ∈ Rⁿ} is the first sample; ^μk is the center of the ^k cluster; ^{r,r_i,k ∈^(0,1)} is the cluster assignment matrix Regression analysis: using linear regression modeling to explore the key factors influencing changes in student thinking, such as course participation and frequency of activity participation; Predictive modeling: using decision tree algorithms to predict future trends in student thought dynamics to inform interventions[17].

4.2

Data analysis and results

The preprocessed data was analyzed according to the following algorithm[18]: Classroom Interaction Data (CI) It was assumed that the experimental group improved (a growing trend) each week and the control group performed flat.

(15)

C I_{i, t} = μ_{C I} + Δ_{E x p} \cdot t + ò, ò ~ N (0, 5)

$$C{I_{i,t}} = {\mu _{CI}} + {\Delta _{Exp}} \cdot t + o,\quad o\sim N(0,5)$$

The experimental group had a baseline value of ^{μCI,Exp = 70} and a weekly lift of ^{Δ_Exp = 2}; the control group had a baseline value of ^{μCI,Ctr = 65} and no lift.

Online Learning Data (OL) Online data correlates with classroom interactions, adding some noise: (16) $O L_{i, t} = α \cdot C I_{i, t} + μ_{O L} + ò, \overset{`}{ò} ~ N (0, 3)$ $$O{L_{i,t}} = \alpha \cdot C{I_{i,t}} + {\mu _{OL}} + o,\quad o\sim N(0,3)$$ Set the scale factor ^{α = 0.8} and the baseline μOL = ⁷⁵.

Questionnaire feedback data (QF) Questionnaire feedback is influenced by a combination of classroom interaction and online learning: (17) $Q F_{i, t} = 3 + 0.01 \cdot C I_{i, t} + 0.01 \cdot O L_{i, t} + ò, \overset{`}{ò} ~ N (0, 0.2)$ $$Q{F_{i,t}} = 3 + 0.01 \cdot C{I_{i,t}} + 0.01 \cdot O{L_{i,t}} + o,\quad o \sim N(0,0.2)$$ Comprehensive performance calculations The overall performance score is based on a weighted formula: (18) $F_{i, t} = 0.4 \cdot C I_{i, t} + 0.35 \cdot O L_{i, t} + 0.25 \cdot Q F_{i, t}$ Example of data fusion The data for a student (experimental group, week 4 as in Figure 8) are as follows:

(19)

(C I_{-} {i, 4} = 78.2), (O L_{-} {i, 4} = 155.6), (Q F_{-} {i, 4} = 4.5)

Overall performance scores:

F_{i, 4} = 0.4 \cdot 78.2 + 0.35 \cdot 155.6 + 0.25 \cdot 4.5 = 31.28 + 54.46 + 1.125 = 86.865

4.3

Data analysis and formula calculation

4.3.1

Comparison of trends between experimental and control groups

Mean value calculation formula(20): (20) $\bar{F} Exp, t = \frac{1}{n} \sum i = 1^{n} F_{i, t}, \bar{F} C t r, t = \frac{1}{n} \sum i = 1^{n} F_{i, t}$ The hypothetical 10-week trend in mean scores for the experimental and control groups is as follows: Table 3.

Mean score control table

Weekly	The mean score of the experimental group	The mean score of the control group
1	70.1	68.9
5	81.5	70.0
10	92.5	71.8

Elevation calculations: (21)

Δ F_{E x p} = \bar{F} E x p, 10 - \bar{F} Exp, 1, Δ F_{C r r} = \bar{F} C t r, 10 - \bar{F} C t r, 1

Results:

Δ F_{E x p} = 92.5 - 70.1 = 22.4, Δ F_{C r r} = 71.8 - 68.9 = 2.9

4.3.2

ANOVA (Analysis of variance)

The experimental[19] and control group data were set up as two independent samples: The mean square error between groups: (22) $M S_{between} = \frac{\sum_{k = 1}^{2} n_{k} {({\bar{x}}_{k} - \bar{x})}^{2}}{g - 1}$ Within-group mean squared error: (23) $M S_{within} = \frac{\sum_{k = 1}^{2} (n_{k} - 1) s_{k}^{2}}{N - g}$ F-statistic: (24) $F = \frac{M S_{between}}{M S_{within}}$ G-By calculation, the hypothesi ^{F = 15.8, p < 0.01_s}, suggests a significant difference between groups.

4.4

Empirical analysis

Trend analysis over time The trend of the mean scores of the experimental and control groups over time was calculated using the formula: (25) $\bar{F} t^{E x p} = \frac{1}{n} \sum i = 1^{n} F_{i, t}, \bar{F} t^{C i r} = \frac{1}{n} \sum i = 1^{n} F_{i, t}$ Example calculation: Week 1 experimental group mean score: $\bar{F} 1^{E x p} = 72.1$ control group: ^{$\bar{F} 1^{C t r} = 70.5$} Week 10 experimental group mean score: $\bar{F} 10^{E x p} = 92.5$ control group: ^{$\bar{F} 10^{c t r} = 72.3$} Visualization is shown in Figure 9: the performance of the experimental group increased significantly over time and the test of significance of the difference between the groups.

A one-way analysis of variance (ANOVA) was used to test the difference in performance between the experimental and control groups[20].

The mean squared error between groups: (26) $M S_{between} = \frac{\sum_{k = 1}^{g} n_{k} {({\bar{x}}_{k} - \bar{x})}^{2}}{g - 1}$ Within-group mean squared error: (27) $M S_{within} = \frac{\sum_{k = 1}^{g} (n_{k} - 1) s_{k}^{2}}{N - g}$ F-statistic: (28) $F = \frac{M S_{between}}{M S_{within}}$ RESULTS: The value for the experimental group was much less than 0.05, indicating a significant difference.

Kalman Filter Dynamic Analysis Smoothing the time-series data of student performance in the experimental group, using the Kalman filter formula(29): (29) $\hat{x} k = \hat{x} k - 1 + K_{k} (z_{k} - {\hat{x}}_{k - 1})$ Assume measurement noise ^{R = 5}, process noise ^{Q = 2}, and gain ^{K_k = 0.7}.

The filtered results show that student performance improves week by week with a smoother trend as shown in Figure 10.

Key Factors Influencing Ideological Characteristics Secondly, social practice participation had the same significant effect (p < 0.01). Students who participated in voluntary activities and social practice showed a stronger tendency to form a Marxist identity. This suggests that social practice not only provides students with the opportunity to contact and understand society, but also allows them to realize the importance of collectivism and social responsibility in practice, thus making it easier for them to identify with the basic concepts and values of Marxism. This finding emphasizes the importance of combining theory and practice in ideological and political education. Finally, the effect of network culture also showed some significance (p < 0.05). Overexposure to entertaining and individualized network culture content may lead to students’ ideological fluctuations and affect the formation of their values and worldviews. Students in the network era are faced with the diversity and complexity of information dissemination, and in the face of massive information, how to effectively filter and internalize meaningful content has become a new challenge for ideological and political education. Therefore, while strengthening ideological and political education, teachers and educators should also pay attention to students’ online cultural consumption behaviors and guide them to establish correct values in order to resist the erosion of negative cultural influences.[21].

5

Results and analysis

Through the clustering analysis of students’ behavioral, cognitive and affective characteristics, this paper divides the student group into three categories, namely, high identity group, medium identity group and low identity group, and finds that there are significant differences in each group in terms of each characteristic, as shown in Figure 9.

High identification group (Cluster 1): this group of students showed high class participation and attendance, with mean values of 92.5% and 10 times/week, respectively, which were significantly higher than the other groups. In addition, online quiz scores and video completion rates were significantly better than the other groups, with mean values of 91 and 95%, respectively. Students in this group not only excelled cognitively, but were particularly active in class discussions, participating in discussions an average of 13 times per month. Affective tendency scores were all positive, indicating that they had a strong sense of identification with the curriculum and the content of ideological and political education, and the scores for values identification and course satisfaction were both 5. The high identification characteristics of this group suggest that they are an ideal audience for the realization of teaching objectives, representing those students who have fully accepted and identified with ideological and political education. Teaching for this group can focus more on depth and expansion to stimulate their higher level thinking and exploration.

Middle Identity Group (Cluster 2): Students in the middle identity group performed at a moderate level in terms of behavior and cognition, with an attendance rate and class participation of 77.5% and 6 times/week, respectively. Online quiz scores and number of discussions were also intermediate, with mean values of 77 and 8 times/month, respectively. Although satisfaction and values identity scores for this group were slightly lower than those of the high identity group, they still showed strong positive tendencies overall, with affective scores mostly centered between 0.3 and 0.5. These students have the potential for further guidance and advancement, but their engagement and identification are relatively weak, and they may need more support in terms of motivation and educational resources to enhance the attractiveness and effectiveness of their ideological and political education.

Low identification group (Cluster 3): students in the low identification group showed lower indicator values in behavioral, cognitive and affective aspects. Attendance and class participation were the lowest at 55% and 3 times/week respectively, which were significantly lower than the other groups. Additionally, online quiz scores averaged only 63 and discussions were extremely sparse at 3 times/month. This group had generally negative affective tendency scores (e.g., -0.3 or -0.5) and satisfaction and values identity scores that were generally below 2, suggesting that they have a weaker sense of agreement with the course content and may be experiencing higher levels of ideological volatility. Focused attention is particularly needed for this group, especially in the promotion of ideological and political education and classroom participation mobilization. Educators should adopt personalized interventions to stimulate their willingness to participate, and gradually improve their sense of identity and participation through enhanced classroom interaction and social practice.

In summary, the differences among the three groups of students in ideological and political education reflect the need for different teaching strategies. The high identity group should be further deepened in education content, the middle identity group needs to be provided with more guidance and support, while the low identity group should be taken as the focus of attention, and their interest and participation should be stimulated through a variety of ways to enhance their sense of ideological identity.

1)

High efficiency of thought dynamics prediction modeling

The decision tree model demonstrates high predictive ability, and the specific results are as follows: the accuracy of the model reaches 93.0%, which is significantly better than the traditional linear regression and logistic regression models in the task of dynamic classification of ideas, and shows a strong classification effect. The precision rate is 91.2% and the recall rate is 92.5%, which indicates that the model has a good balance in identifying different categories of students and can effectively distinguish between high identity, medium identity and low identity groups.The F1 value is 91.8%, which further validates the overall performance of the model in the classification task. In the feature importance analysis, attendance (30%) and online quiz score (25%) were rated as the most important predictors, emphasizing the key role of students’ behavioral and cognitive characteristics in the analysis of thought dynamics. Affective tendency scores (20%), on the other hand, provided a valid measure of students’ intrinsic attitudes and values, especially when identifying the low-identification group, which showed a high weight, suggesting that affective factors could not be ignored in their influence on ideological identification. Despite the lower weights of video completion rate and class participation, they are still informative in predicting students’ engagement in learning and class participation [23].

2)

Scientific nature of educational resource optimization

Optimizing the allocation of resources through linear programming methods, the results of the study show that balancing fairness and efficiency is the key to achieving educational goals when resources are limited:

The high identity group allocates 20% of the resources for consolidating existing positive attitudes and behaviors;

The medium-identification group was allocated 30% of the resources to motivate them to move to a higher state of identification;

the low identification group allocates 50% of its resources to maximize the level of ideological identification.

By adjusting the resource allocation, the low identification group had the most significant expected improvement in thought dynamics (from 63 to 78 points), while the high identification group had the smallest improvement, but also the highest cost-benefit ratio.

3)

Innovation of data fusion method

The study adopted a multi-source data fusion technique [24], combining behavioral data (e.g., attendance, class participation), cognitive data (e.g., online quiz scores, frequency of discussion), and affective data (e.g., values identification and affective ratings), and data fusion improves a comprehensive understanding of ideological dynamics [25] and avoids the one-sidedness of a single data source; the feature dimensionality reduction retains 95% of the variance information, ensuring the model performance while reducing complexity; the integration of multidimensional data provides data support for personalized decision-making in future ideological and political education.

6

Conclusion

Based on the Marxist idea of the data fusion model, this paper draws the following conclusions through the weighted average method that can synthesize data from different sources; the Kalman filter method that can update the prediction of educational effects in real-time; the decision tree algorithm that helps educators discover the potential flaws in the data and then make more scientific educational decisions; and analyzing the data and combining it with the ideological education policy of colleges and universities:

The educational intervention strategy of data integration can effectively stimulate students’ learning motivation and improve comprehensive performance scores.

Classroom interaction dimension: the number of questions asked and the discussion participation rate of students in the experimental group increased steadily over the 10 weeks, with the mean value increasing from 70.1 (week 1) to 92.5 (week 10).

Online learning dimension: the experimental group’s average viewing time increased by 23%, and the course completion rate grew from 72% to 96%.

Questionnaire feedback dimension: the mean satisfaction score of the experimental group increased from 3.5 to 4.7, indicating a significant increase in student satisfaction with the content.

The traditional teaching model lacking data-driven interventions makes it difficult to achieve significant learning improvement.

The indicators in the control group showed small changes and a smooth trend. For example, the classroom interaction score increased only slightly from 68.9 (week 1) to 71.8 (week 10).

The online learning completion rate was almost unchanged during the 10 weeks, remaining at about 73% on average.

Interventions based on the integration of classroom interaction and online learning data have significant effects and can be generalized for large-scale educational scenarios. Statistical tests of analysis of variance (ANOVA) and regression modeling showed that the educational intervention (experimental group) was highly correlated with academic achievement improvement, ( p < 0.01), with a significant effect value.

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

Research on the Combination Strategy of Marxist Ideology and Ideological and Political Education in Colleges and Universities Based on Data Fusion Modeling

Zhiqin Zhang

Data publikacji: 27 lut 2025

Otrzymano: 25 wrz 2024

Przyjęty: 15 sty 2025

DOI: https://doi.org/10.2478/amns-2025-0112

Słowa kluczoweWeighted Average Method, Kalman Filter, Decision Tree, Data Fusion Model, Ideological and Political Education

© 2025 Zhiqin Zhang, published by Sciendo

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

Słowa kluczowe
Weighted Average Method, Kalman Filter, Decision Tree, Data Fusion Model, Ideological and Political Education