Driven by big data, e-commerce platforms have accumulated massive user behavior data, which can be transformed into valuable information after cleaning and feature selection. This study analyzes users’ historical behavioral data on e-commerce platforms, constructs a gradient boosting decision tree prediction model based on user, product, category, and two-by-two interaction behavioral features, and extracts designed feature data from raw CSV data based on Hiv as the prediction basis of the model. At the same time, clustering analysis is performed based on the user’s purchasing behavior (dwell time, browsing frequency) to generate user profiles. The experimental results show that after 7 days, the purchase conversion rate of browsing, collecting, adding to cart and purchasing tends to 0. Therefore, the time window for purchase behavior prediction is chosen to be 7 days. In this paper, the prediction model is only trained to 20 epochs, and the Loss value converges to about 0.14, which shows a good training effect. The model has the best classification performance for user purchase behavior prediction, with precision, recall, and F1 values between 0.91 and 0.97. The clustering algorithm divides the user purchase behavior into four clusters, where cluster class 4 has the best user value. In summary, using the gradient boosting decision tree model, e-commerce platforms can more accurately predict user purchasing behavior, thus improving user experience and platform economic benefits.

The convergence of constitutional fundamental rights and administrative enforcement power should pursue multiple legal values, which requires that the operation of the power therein should be more division of labor than cooperation, and that constraints and synergies should be given equal importance. The article will construct the basic constitutional rights and administrative power into the constitutional construction of the subject and the subject of the executive branch, the introduction of the evolution of the game theory to construct the constitutional construction of the subject and the subject of the executive branch of the evolution of the game model, and the design of the game model of the gain function, the replication of dynamic equations and ESS equilibrium point. The initial value of each parameter in the evolutionary game model is set, and the evolutionary stable point of administrative power is simulated by MATLAB software, and the influence of the reward and punishment allocation coefficients on the evolutionary results of the system is explored. When the system evolution stable point strategy is (0,0) and (1,1), the two sides of the game tend to the stable equilibrium state of active cooperation, strengthened regulation and strict supervision. When the reward distribution coefficient and the punishment distribution coefficient gradually increase, the two sides in the evolutionary game system tends to stabilize the point (1,1) the faster the rate will be. In the process of constructing the fundamental rights of the constitution, combining internal and external with the administrative rights list monitoring mechanism can realize the optimal restriction on the application of administrative rights and promote the orderly and stable operation of administrative power.

This topic is based on the perspective of diagnostic evaluation, formative new evaluation, summative evaluation, in-depth analysis of deep learning to help the intelligent development of ideological and political education. The initially formulated questionnaire was modified several times, and the questionnaire design task was finally completed, and the formal distribution of questionnaires began to obtain the data for this study. At this level, the empirical research method combining quantitative research and qualitative research is used to deeply analyze the current situation of the intelligent development of ideological and political education in colleges and universities under the perspective of deep learning, with a view to contributing to the intelligent development of ideological and political education in colleges and universities. The mean value of the five dimensions of ideological and political education to stimulate the effectiveness of active learning (A1), promote the degree of in-depth understanding (A2), deepen the effect of interactive participation (A3), enhance the ability of higher-order thinking (A4), and expand the quality of the transfer of the use of the quality of the five dimensions (A5) is higher than 3.55 points, in addition to the significant difference in the characteristics of the samples (gender, academic qualifications, major, political profile), and puts forward five paths for the development of the five development paths, which are aimed at helping to promote the new era of the intelligent development of Civic and Political Education.

Intangible cultural heritage is an important part of national culture, carrying rich historical information and cultural value. This paper mainly generates digital media art based on the characteristics of intangible cultural symbols through fractal geometry, and builds a digital media communication mode of intangible cultural art on the basis of dynamic texture, so as to realize the digital protection and inheritance of intangible cultural heritage. On the basis of Transformer network model, further combined with fractal geometry technology, inductive bias lacking in Transformer is introduced from the perspective of translation invariance and locality, and the dynamic texture generation method of digital media art combined with fractal geometry and Transformer model is formed in this paper. Experimental results show that the application of translation invariance and locality can increase the ODS, OIS and AP indexes by 17.17%, 27.72% and 25.38%, respectively. In the self-built Miao pattern data set, this method can generate dynamic texture features of Miao culture and art more completely and clearly. At the same time, this paper can create digital media art for Miao embroidery patterns through the generated results of this method, and improve the audience influence and satisfaction of intangible cultural heritage.

Optimization problems usually involve multiple objectives, while fuzzy cognitive maps can effectively show the causal relationship between concepts, and the combination of the two can greatly advance the development of the education field. In this paper, we design a fuzzy cognition-based knowledge map for labor education courses and a multi-objective optimization model for labor education courses to optimize learners’ learning paths and recommend personalized exercises from multiple stages. Through teaching experiments and regression analysis, the teaching effect of the multi-objective optimization algorithm in labor education courses is evaluated. This paper borrows the k-means algorithm to classify learners into four clusters, and the algorithm provides learning path optimization for different clusters of learners in labor education courses. The exercise recommendation accuracy of this paper’s algorithm ranges from 0.91 to 0.97 and has better novelty and diversity recommendation performance. In the experimental class in the fuzzy cognitively oriented multi-objective optimization labor course, the learners’ labor scores improve faster and are about 3.8 points higher than those of the traditional teaching, and the regression results show that this paper’s model has a positive and positive effect on the teaching effect. The average satisfaction scores of this paper’s model in labor education courses for the friendliness of teaching aid, effectiveness of cognitive diagnosis method, usefulness of path optimization, and reasonableness of personalized recommendation of exercises are above 4.3, indicating that the model has practical application value in labor education courses.

As people’s demand for high-quality development of education becomes stronger and stronger, the field of education is paying more and more attention to the fundamental task of education by establishing moral values. In this paper, the improved genetic algorithm based on predation strategy is applied to the learning path recommendation system to realize the auxiliary teaching of Civics education. The study first proposes a Bayesian knowledge tracking model based on multiple interactions for knowledge tracking of students’ Civics and Political Science competence, and carries out model comparison and knowledge tracking visualization and analysis on three datasets and the real dataset of practice questions. Then according to the constructed learner model and knowledge connectivity state model, the personalized learning path construction model is designed by using learner features, knowledge point features, and generic learning paths as inputs, combined with the improved genetic algorithm based on predation strategy. The intelligent assisted teaching system designed in this paper is put into practice for Civics teaching and scored by questionnaire and paired t-test method. The results of the study said that the knowledge tracking model proposed in this paper compared with other models, the model in this paper improves the accuracy rate by 1%~2%. Using the non-elite individual set to enrich the population diversity to participate in genetic operation and iteration, the experiment shows that PSGA performs well in multiple comparisons with PSO and SGA methods, and can construct personalized learning paths more accurately, stably and effectively. The results of teaching practice show that the teaching system proposed in this paper can effectively improve students’ learning ability in Civics.

The further deepening of education informatization has led to a significant shift in teaching methods as well as learning tools, and it is of research significance to explore how to use online learning platforms more effectively in non-traditional teaching environments. In this study, after pre-processing the online teaching data of Marxist theory in the Civics course, the Squeeze method is used to extract the relevant features of teaching interaction behavior in the data. Convolutional neural network is used to realize the prediction of teaching interaction behavior based on the input features, so as to realize the real-time intervention and effect enhancement strategy of teaching interaction. It is verified that the ICAM-ResNet neural network prediction model proposed in this paper has a good effect in making online teaching interactive behavior prediction, and the prediction accuracy can reach 0.816. After implementing the intervention strategy according to the prediction results, the average online learning time of students increased from 30.61 min (1 class period) to 44.54 min (16 class periods), and most of the students would actively answer the questions in the classroom, and the rate of answering correctly increased, so that the effect of teacher-student interactions was substantially improved. On the one hand, this study provides a new way of thinking for the teaching research of Marxist theory course, on the other hand, the results of the study are conducive to optimizing the teaching practice of the course and promoting the teaching interaction, so as to promote the development of the teaching of the course.

The concept of urban green development promotes the development of intelligent technology as a new energy power, so that digital intelligent technology continues to enter into people’s vision, but also gradually accepted by the people. However, there is a lack of research on intelligent perception that focuses on residents’ attitudes, so this paper takes the theory of perceived value as the basis to analyze the influence path of intelligent perception of urban green space. Based on structural equation modeling, this paper explores the relationship between intelligent perception and residents’ attitudes in terms of perceived functional value, perceived emotional value, perceived social value, cognitive value and perceived risk. Then the intelligent perception prediction model for urban green space is constructed by using variational modal decomposition (VMD) combined with support vector regression (SVR), and the actual performance of this paper’s model is examined through experiments. This paper takes City Y as an example for prediction, and the results show that the intelligent perception of green space in City Y from 2023 to 2026 continues to show an upward trend. In addition, in order to prove the superiority of this paper’s model, its MAE, MAPE, RMSE and IA are compared with the prediction models of ARMA, BP, SVR and RF, respectively, and this paper’s model achieves the best results with the values of 4.2495, 15.8082, 3.5247 and 0.5225 for each index. In conclusion, the prediction model proposed in this paper has high accuracy in intelligent perception prediction.

The big data environment is dynamically changing, so the multi-objective optimization algorithm for the integration of English translation information technology needs to have dynamic adaptability. In this paper, we first construct a multi-objective learning parameter model for English translation information technology. Then a reference point-based environment unpredictable dynamic multi-objective optimization algorithm (UDERP) is proposed to realize the dynamic adaptability of the multi-objective optimization algorithm. Finally, the designed English translation information technology incorporating the UDERP algorithm is simulated and tested. The performance of UDERP algorithm, DNSGA-II algorithm and DSS algorithm are compared with each other using three test functions of FDA series. When the environment changes the optimal solution derived from the algorithm proposed in this paper is closer to the real Pareto solution. Comparing the neural machine translation based on cross-language pre-trained language model and the neural machine translation based on multi-coverage model, the English translation information technology designed in this paper has a better convergence effect and can realize more accurate parameter estimation.

Fintech has not only greatly improved the operational efficiency of banks by introducing cutting-edge technologies such as big data, artificial intelligence, and blockchain, but also posed new challenges to banks’ risk management. This paper uses Monte Carlo simulation to explore the impact of fintech on banks’ operational efficiency and risk management. A VaR data model is used to analyze the impact of fintech on the operational efficiency of three types of commercial banks: big five banks, joint-stock banks, and city commercial banks. The non-performing loan ratio of China MS Bank is used as the empirical object for quantitative analysis of bank risk management. Monte Carlo simulation is used to realize the VaR calculation of banks’ NPL ratio. The empirical analysis finds that the impact of fintech on the operational efficiency of all three types of commercial banks is relatively significant, but there are differences in direction and lag period. Meanwhile, FinTech increases banks’ NPL ratio. It shows that fintech has a negative impact on bank risk management, for this reason, this paper develops relevant strategies to deal with the risk challenges brought by fintech according to bank types.