Based on the demand of load balancing in distributed system scenarios, this paper introduces the concept of dynamic priority in the algorithm and designs the dynamic feedback load balancing (DFLB) algorithm for numerical analysis. Through the closed-loop process of collection-feedback-utilization-collection, the overall performance of the system is realized. The Mininet tool and the Floodlight controller are used when building the load balancing system experimental environment to verify the reliability of the algorithm from the response delay, throughput and other indicators. The study shows that the DFLB algorithm reduces the response time of the system by about 20% compared with the static deployment method, and the DFLB algorithm reduces the load variance, saves computational resources, and makes the load of the system more balanced and efficient. The average throughput of the DFLB algorithm is improved by about 10% compared with the PALB algorithm and DALB algorithm, and 6% compared with the PALB algorithm and DALB algorithm, respectively. Starting from 1000 concurrent connections, the DFLB algorithm has a higher access rate. Thus, the algorithm leads to an improvement in the overall performance of the system.

The field of machine translation has made significant progress in recent years, but how to improve translation accuracy and context consistency is still an urgent challenge. In this paper, a context-aware translation accuracy improvement strategy based on deep reinforcement learning is proposed for English translation. Based on CNNs neural machine translation model, the multi-intelligence deterministic deep policy gradient algorithm is utilized to combine the output of the translation model with the human evaluation index (BLEU), and the reward function is constructed to guide the model learning. In addition, in order to enhance the context-awareness of the model, the study introduces a context encoder in the deep reinforcement learning framework to capture sentence-level contextual information and incorporate it into the translation process. The experimental results show that the optimized model has better training performance, with 40 epochs of iterations, the Loss converges to 0.135 up and down, and its English translation F1 value is 94.95%. And as the number of encoder layers rises, the number of semantic high-level features increases. The N-GRR difference between the generated translation and the standard translation of the model in this paper is the smallest, and the over-translation phenomenon is less. The number of out-of-set word interference is more than 6, and the BLEU value of this paper’s model is improved by 17.89% to 55.55% compared with the comparison model. And the algorithm has good translation performance, with METEOR scores of 0.562~0.803 on different topics. The research results fully verify the effectiveness of deep reinforcement learning based on deep reinforcement learning to improve the accuracy of English machine translation.

This paper proposes a risk indicator system for mental health management of college students that takes individual developmental status, social environment, human-computer interaction, and negative emotions as the first-level indicators, and clarifies the path of obtaining mental health management monitoring data, the weights of the indicators, and the safety warning interval of mental health management. Because of the uncertainties in the mental health management of college students, fuzzy logic is introduced to deal with the uncertainties of environmental changes, student behavior and other factors in the mental health management, and to improve the level of mental health management in colleges and universities. A fuzzy logic-based risk warning model for mental health management of college students is designed. The mental health status of students is further refined by the SCL-90 scale, and the mean score level of each factor of the scale is compared with the youth norm and adult norm. Input the fuzzified student mental health data in the fuzzy logic risk early warning model, and output the risk score of the fuzzy logic model for mental health management of college students. When the set threshold is 60, the fuzzy logic risk early warning model can effectively identify the abnormal values of students’ mental health, and the early warning model has practical utility.

How to communicate with users in a timely and effective manner and determine the intentional purpose of customers plays an important role in promoting continuous user interaction and improving service efficiency in the power marketing industry. The article firstly researches on a single-round natural language understanding algorithm based on intent-slot bi-directional interaction, which adopts a bi-directional information flow to realize the bi-directional information interaction between intent and slot. In the intention recognition layer, the interaction attention mechanism is utilized to introduce slot context information. Then the overall design scheme for the construction of an intelligent customer service system for power marketing from dialogue state keeping, multi-round question and answer, model storage to answer visualization is proposed, and the potential functional requirements are analyzed exhaustively. Finally, experiments from various aspects prove the effectiveness of the proposal in this paper. In the comparison experiments on MixATIS with MixSNIPS dataset and DSTC4 dataset, the metrics are improved by 0.3%, 1.5% and 0.5% respectively when comparing GL-GIN model on MixATIS dataset. This leads to the feasibility of the intelligent customer service system for power marketing constructed in this paper.

In recent years, socio-economic development and the process of massification of vocational education have been accelerating. The article surveys the current situation of the articulation between vocational education and undergraduate education through questionnaires. On this basis, in order to better realize the cultivation of employment-oriented talents, it designs a teaching resource acquisition method based on computational optimization, constructs a crawler search method by fusing genetic algorithm and ant colony algorithm, and realizes automatic clustering by using a clustering algorithm based on the combination of K-mean and particle swarm algorithm in random search direction. The results show that only 23.3% of the students think that there is no duplication of content between vocational and undergraduate education, 89.6% of the students want to set the teaching content according to different needs, and the current talent cultivation for the articulation of vocational and undergraduate education suffers from poor wholeness and monotonous tendency. The proposed crawler search method and automatic clustering method show superior performance and can accurately extract teaching resources and process structured information. Finally, the employment-oriented talent cultivation model is proposed to actively explore the path of integrating vocational and undergraduate education and promote the development of vocational education.

Key frame extraction is an important research content for human motion capture data analysis and processing, for this reason, a key frame extraction method for motion capture data based on quantum particle swarm optimization algorithm is proposed, which can either extract a definite number of key frame sequences or extract key frame sequences according to the objective function. In this paper, the spatio-temporal graph convolutional network is selected as the benchmark network for tap dance action recognition, and the dance action recognition is realized by combining adaptive and attention mechanisms. The comprehensive index of tap dance is introduced and used as a constraint, and the golden section algorithm is used to optimize the training path of the dance action to obtain an ergonomic training path. The experimental results of this paper show that the key frame extraction method of motion capture data based on quantum particle swarm optimization algorithm meets the need of real-time compression of motion capture data. By constructing the validation dataset, the accuracy improvement of AAST-GAN algorithm and the effect of gesture extraction are compared and verified, and the recognition accuracy reaches more than 86%, which is a good recognition accuracy for each tap dance action. The dance movement training path proposed in this paper ensures the effectiveness and comfort of tap dance movements.

Dance Anatomy is a basic theory course for university dance majors, which reveals the structure and function of various parts of the human body and their important roles in dance training through an in-depth interpretation of dance anatomy. Using relevant equipment and instruments, we will set up a data acquisition environment for data acquisition and pre-processing. For the problem of coordinating music rhythm and dance movement, a time-series autoregressive model is used to realize music-driven dance synthesis, and the model loss function is clarified. Combining the above model, data, and modeling software, the task of modeling the human dance movement mechanism is completed, and the cosine similarity is adopted to analyze the problem of coordinating music rhythm and dance movement. In both the training and test sets, the music-driven dance sequences and the original sequences fluctuate within a certain range (-8, 13), and the scoreRatio value of this paper’s method (1.505) is much better than that of the other four sets of models, which verifies the efficacy of its model in the application of the task of modeling the mechanism of human dance movement, and also verifies the reliability of cosine similarity method. This will enable better implementation of human movement mechanisms in dance anatomy into practical scenarios, help trainers to better perform dance training and performance, reduce dance injuries and prevent occupational diseases.

Aiming at the many problems in research resource management in private universities, this paper takes the integration of research resources in international business discipline of Xiamen Institute of Technology as an example, proposes a global integration and dynamic allocation model of research resources in distributed computing environment based on mobile agent (DCMA), and designs a dynamic bidirectional matching method of tasks and resources (DBMM) in order to improve the effectiveness of distributed computing. Experiments show that the proposed DBMM algorithm outperforms the LDCP algorithm and the hierarchical node sorting algorithm (SNLDD) in three metrics, namely, scheduling length, acceleration ratio and computational efficiency. Compared with LDCP and SNLDD, the scheduling length of DBMM algorithm is shortened by an average of 19.89% and 11.81%, the acceleration ratio is improved by an average of 19.77% and 9.26%, and the computational efficiency is increased by an average of 10.74% and 3.72%, which further improves the resource utilization rate of distributed computing system. Experiments were conducted using the research resource integration model, which achieved better efficacy in terms of probability value, goodness-of-fit, and stability of research resource integration in international business disciplines compared with the gray correlation analysis method. This paper provides an example reference for distributed computing system to realize research resource integration and efficiency improvement.

Aiming at the dilemma of corpus-based intelligent English translation, the article proposes an English neural machine translation method based on depth-separable convolution, which combines with the dynamic computation method to improve the semantic consistency of the translation system for semantic alignment and fusion. In order to verify the training effect of the proposed convolutional neural network model combined with the dynamic computation method, comparison experiments with one-way and two-way network models and baseline model with different cut-off granularity are conducted respectively. In order to better examine its performance in practical translation applications, online translation, machine translation and systematic methods are utilized for comparison. The BLUE values of this paper’s model for Chinese-English data translation in four different granularities of words, syllables, subwords and characters are 21.41%, 21.91, 29.25% and 20.40%, respectively. In 100,000, 200,000 and 500,000 training English-Chinese bilingual parallel corpus, the training time consumed by the model in this paper is 9.58 h, 15.94 h and 32.69 h. In practical application, the decibel range of the noise reduction of the translation system method designed by the research is distributed in [1.62 ~ 1.89], the average value of coherence is 91.1%, and the average compression rate and the average stability of the BLEU scores are 93.84% and 98.38%, respectively, and the results are better than the comparison methods.

This paper constructs a set of models for monitoring and evaluating the effect of Civics education through the research on the evaluation of Civics education based on educational big data environment. First, based on distributed gray cluster analysis, it analyzes and researches students’ Civics learning behavior, and explores learners’ learning characteristics by mining meaningful behavioral features for cluster analysis. The second is to design the Civics teaching quality evaluation model using principal component analysis, test the effects of population size and convolution kernel number on the performance of the Civics teaching quality evaluation model, and optimize the teaching quality evaluation model by using the dimensionality-reduced evaluation data. Distributed gray cluster analysis gets four clusters according to the characteristics of students’ learning behaviors, which are divided into excellent, diligent, average, and negative students.PCA selection of evaluation indexes found that the cumulative contribution rate of the first 10 principal component indexes to the evaluation of the quality of Civic Teaching in colleges and universities has reached 95.63%, which indicates that these 10 indexes can adequately evaluate the quality of Civic Teaching in colleges and universities. When the number of population size is taken as 31 and the number of optimal convolution kernels is taken as 19 values, the RMSE of the evaluation model is 0.01973, and the test time consumed is 0.0783ms, which is the best performance. The constructed Civics education effect monitoring model can effectively assess students’ learning behavior and efficiently and accurately evaluate the quality of Civics teaching.