As a key component of urban environmental resources, the design of landscape paths and facility layouts of urban public environments is not only related to the overall aesthetics of the city, but also to the quality of life of urban residents. In this paper, from the perspective of landscape layout, the ecological landscape spatial network is constructed by calculating the ecological landscape environmental adaptation degree and the ecological landscape pattern index. On this basis, the traditional ant colony algorithm is introduced and its heuristic function and path selection are improved, and the adaptive adjustment factor and angle guiding factor are added to improve the diversity and efficiency of path searching, so that the landscape layout optimization model based on the ant colony algorithm is obtained. Using this model to design a landscape layout optimization scheme for a scenic spot, the average fulfillment time of the optimized landscape path is 20.73 minutes, which is 19.52 minutes shorter than the average fulfillment time of the original planning scheme, indicating that the model in this paper is able to carry out the landscape layout optimization design effectively.

ECG and PCG reflect the activity characteristics of the heart, and the combination of the two can record the electromechanical activity information of the heart more comprehensively. In this paper, we design a heart failure prediction model based on Transformer, and utilize Transformer Encoder to complete the feature fusion of ECG and PCG. Feature classification is performed using ResNet-18 to achieve the prediction of nine typical arrhythmias. Evaluate the classification results on the dataset to explore the performance level of the proposed model. Obtain ECG and PCG data in real situations, and select entropy analysis and heart rate variability metrics to quantify the physiological signal time series complexity. The model classification accuracy, specificity and sensitivity are compared to analyze the effect and superiority of the proposed model in practical applications. The results show that the average accuracy of the model on the four datasets reaches 92.28%, and the highest average F1 score is 0.930. In practical applications, the classification accuracy, specificity and sensitivity of the proposed model in this paper are 96.79%, 97.47% and 96.77%, respectively. Through the fusion analysis of ECG signal and heart sound signal characteristics, the model fully reflects the HRV change characteristics of heart failure patients and can effectively predict heart failure.

National security education in the new era puts forward new and higher expectations on the scope, degree, speed, and object of knowledge dissemination, while presenting new dissemination characteristics such as all-media and group emergence.Based on graph theory algorithm, this study proposes a dissemination model with credibility constraints about national security education knowledge.Text mining is used to analyze discussions of social network users on national security education knowledge from Sina Weibo and Baidu Search. The dissemination mechanism of national security knowledge is explored through text analysis. Based on this, different expectations of information dissemination are set to conduct numerical simulation. The simulation results show the model is highly sensitive to parameter changes. In the case of R < 1, with the increase of β, the time for S to reach the steady state decreases, and the time for I to reach the maximum value decreases, while the maximum value increases.When β = 0.03, Max I = 39.86; and when μ = 0.3, Max I = 37.23. The model plays an important role in controlling and managing knowledge dissemination.The proposed graph theory-based knowledge diffusion model achieves an average knowledge stock of 0.924 under regular networks and 0.726 under scale-free networks. In terms of knowledge diffusion rate, this model outperforms both the traditional knowledge diffusion model and the random diffusion model.

In this paper, DAG is utilized to represent the dependencies between musical features, and a topological sorting algorithm based on layer order relationships is used as the sampling algorithm for AI music generation models. The feature de-entanglement mechanism of VAE is utilized to learn multiple feature representations, and Transformer-XL is used as the encoder and decoder of the model to design the Control-VAE model that manipulates the latent variable representations to change the music structure. Statistical autocorrelation coefficients, spectral analysis, and diversity auto assessment metrics data were used to evaluate the model performance in terms of three dimensions: melody, timbre, and diversity. The feasibility of Control-VAE model AI music generation and melody optimization is examined through the evaluation of practical application effects. The results show that the autocorrelation coefficients and frequency amplitudes of the music generated by Control-VAE model are basically consistent with the original music, and reach human-like PPL values, seIf-BLEU values and Zipf coefficients near p=0.95.The music pieces generated by Control-VAE model have a certain degree of musicality, and the melody-optimized music is clear, accurate and novel and interesting.

Based on the background of information technology, this paper proposes a multimodal blended learning model of English listening based on “WeChat+Classroom+TED-Ed”. It focuses on the experimental teaching of multimodal learning and English listening comprehension, and describes the object of the study, the design of the study and the process of the study. Based on the research idea, the experimental variables were designed, and the empirical analysis was carried out by using multiple linear regression model. The teaching effect of multimodal teaching is examined by comparing the differences in the total English listening scores of the two groups of students before and after the experiment. With the help of Pearson correlation analysis, the correlation between the experimental variables is explored. The value of R² was determined through the multiple regression model to determine the magnitude of the explanatory power of multimodal learning on English listening comprehension ability. The results showed that the scores of the control class improved by 1.19 points and the experimental class improved by 4.19 points in the experimental posttest, with a significance (two-tailed) p-value = 0.008<0.05. The explanatory power of the combined three modalities of learning on English listening performance was 15.4%, and classroom learning had the highest level of significance in terms of its explanatory power on listening comprehension, and the test of regression coefficients reached the level of significance (t=3.862, p= 0.002<0.05).

As artiϐicial intelligence technology becomes more and more mature, it is both a challenge and an opportunity for English speaking teaching. Aiming at the poor generation of virtual English teaching resources due to the training problems of traditional generative adversarial network, dual generative adversarial network is used to optimize the above problems and select the virtual English teaching resources that meet the requirements with the help of Pielou. At this level, the HTC VIVE suite, high performance computer system, Unity 3D development engine, and joystick control are integrated to jointly complete the work of English speaking teaching scene design. Combining the research data and evaluation indexes, the practical application efϐicacy of the scenario is analyzed. From the overall performance of different methods in the four datasets, this paper’s method is superior to the other four methods, that is, this paper’s method is able to generate high-quality virtual spoken English teaching resources. And the practical application efϐicacy in terms of test scores, learning effects, satisfaction, and English speaking teaching background is better than traditional multimedia, which is more conducive to promoting the development of English speaking teaching.

In order to more comprehensively study the influencing role mechanism of consumer behavioral decision-making process in the digital economy platform and explore the influencing factors of consumer behavioral decision-making, this paper constructs a model of consumer behavioral decision-making process based on Bayesian network. With the help of Netica software to construct the Bayesian network topology, using EM algorithm to learn the parameters of the Bayesian network model, and proposed to use the Bayesian network to carry out sensitivity analysis and probabilistic inference, and formulate the corresponding Bayesian network model framework. Subsequently, the influencing factors of channel search willingness and purchase willingness and their relationships in the consumer behavioral decision-making process in the digital economy platform environment are analyzed. The structural equation model is introduced, the measurement equation and structural sub equation calculation methods are determined, and the sample data are collected by means of questionnaires to carry out the test and analysis of the model of consumer behavioral decision-making process. The CR value of each variable in the model of this paper is higher than 0.7, and the AVE values are all greater than 0.5, and the model performs well in terms of intrinsic quality. The exogenous latent variables such as perceived benefits, channel trust, and transfer costs have a significant positive effect relationship on the endogenous latent variables such as search behavior and purchase intention (P<0.05).

Dress metaphor is a very important way of expression in the novel text of Ming Dynasty, and the recognition and interpretation of the metaphor play a very important role in really understanding the novel text. This paper proposes a dress metaphor recognition model based on Transformer and graph convolutional neural network, and a dress metaphor interpretation method based on Seq2seq framework. The apparel metaphor recognition model performs feature extraction of global and local information of apparel metaphor sentences by Transformer. Graph Convolutional Neural Network is utilized to obtain syntactic structure information and sentence dependencies, in order to complete multi-word dress metaphor recognition. Then the obtained deep metaphor features and syntactic structure information of the sentence are input to the classification layer. The metaphor decoding method carries out costume metaphor understanding through the encoder-decoder, which chooses the LSTM network structure for both encoder and decoder to better obtain the semantic features of the novel text. The dress metaphor recognition model improved the recognition correctness on the dataset by 17.97% and 7.28%. The dress metaphor interpretation method based on the Seq2seq framework elaborates the interpretation content and can more accurately interpret the dress metaphors in Ming Dynasty novels. It verifies the practicality of the metaphor recognition and interpretation model in this paper in the task of interpreting dress metaphors in Ming Dynasty novel texts.

The higher the corporate financial transparency, the more it can reduce the information asymmetry, which can enhance the market trust and improve the corporate performance. In order to improve corporate financial transparency, the study constructs a financial fraud identification model by improving the machine learning model based on XG Boost algorithm from the financial fraud factors. Based on the XG Boost algorithm, the model integrates the decision rules through the weighted fusion method to generate a new decision tree to determine the financial fraud. In order to improve the ability of enterprise performance assessment, the baryon support vector machine method is used to classify the performance of enterprise employees, and the nonlinear baryon support vector machine is used to establish the enterprise performance assessment model. In the process of verifying the effect of the two models, text indicators are extracted using big data technology to provide a rich feature set for the financial fraud identification model. The data from ERP, CRM and other systems are integrated to provide a comprehensive and high-quality data set for the enterprise performance assessment model. After empirical analysis, the combination of big data and machine learning can improve the effect of financial fraud identification, and then effectively improve the transparency of corporate finance. The enterprise performance evaluation model provides a scientific and efficient quantitative evaluation tool for enterprise managers, and effectively improves the enterprise performance evaluation capability.

The integration and development of Sichuan’s rural music and cultural tourism industry is of great signiϐicance in the context of rural revitalization strategy. The purpose of this paper is to construct a multilevel regression model to deeply explore the inϐluencing factors and role mechanisms of the integration of the two. Through theoretical analysis and empirical research, the research variables are clariϐied, and the null model, random effect model and complete model are constructed and data validation and analysis are carried out. The results show that the richness of rural music resources, the level of cultural and tourism industry, policy guidance and support, market demand and human resources have a signiϐicant positive impact on the integration of rural music and cultural and tourism industry in Sichuan. The results of the full multilevel regression model show that the same level of rural music resource abundance has different impacts on the integration of rural music and cultural and tourism industries due to regional differences. The results of the study provide theoretical support for the development of cultural tourism industry in Sichuan Province, and deeply help the implementation of rural revitalization strategy in Sichuan Province.