Heilongjiang International University

This paper focuses on the issues related to price discrimination under the background of big data, based on the actual situation and problem of Chinese Internet users' consumption data, through the methods of literature research and case analysis, and theoretically expounds its economic essence, so as to find the existing economic problems under the background of big data. The research background and significance of this paper point out that modern enterprises use big data to collect massive users' Internet browsing history and other actual situations to achieve precise marketing plan of enterprises, which makes the hot topic of "big data killing" cause a lot of disputes among users. It can be seen that discussing this issue is of great significance to promote the legal use of personal information records by enterprises, and protect the rights and interests of consumers. Then it analyzes the impact of big data on price discrimination, including the specific implementation path of "data-algorithm" and further makes a specific comparison with the traditional sense of price discrimination. This study also points out the damage of price discrimination to consumer experience and consumption fairness caused by the current situation, and there is no corresponding sound system, and some Internet industries still hide the defects of black box algorithm, and other specific situations, which urgently need to put forward solutions. In view of the shortcomings of regulation and algorithm transparency, this paper describes the specific content and implementation methods of China's digital economy in the future.

Dysregulation of PEAK1 (pseudopodium-enriched atypical kinase 1) plays a critical role in various cellular processes, including cell migration, proliferation and survival. Its aberrant expression or activity has been implicated in the pathogenesis of several diseases, particularly cancer. Early detection allows for the development and application of targeted therapies that can inhibit PEAK1 activity, potentially improving treatment outcomes. This study provides a new method for early diagnosis of PEAK1 by utilizing the specific RNA cleavage ability of Cas13a combined with electrochemical sensing technology. CRISPR/Cas13 has cis cleavage activity and can specifically recognize and cleave target RNA. Subsequently, its trans cleavage activity is activated to non-specifically cleave other single stranded RNAs, resulting in detectable signal changes. In the experiments conducted, high sensitivity for detecting PEAK1 mRNA was achieved by optimizing the interface cleavage of the hairpin reporter probe (ReRNA) molecule. The linear detection range is from 1 pg μL−1 to 10 ng μL−1, with a detection limit of 0.45 pg μL−1. In addition, the results showed that the developed biosensor has good repeatability, reproducibility, and stability, which provides a novel method for the early screening of PEAK1.

The rapid development of artificial intelligence and intelligent algorithms is profoundly transforming the field of education, promoting the widespread deployment of language translation technology in teaching applications. In recent years, bilingual flipped classrooms have raised higher requirements for semantic modeling and ambiguity handling in translation systems. To improve the accuracy of Chinese Japanese translation in complex contexts, a neural network translation framework based on multi input fusion mechanism and iterative back translation strategy was developed. This method introduces contextual cues and ambiguous phrase information as auxiliary inputs based on Transformer, and combines bidirectional pseudo corpus generation and quality filtering mechanisms to achieve semantic enhancement of the Sino Japanese bidirectional translation system through continuous alternating training. The experimental results show that the proposed model has an accuracy of 0.94 in Test 1, a root mean square error of only 0.21, and a BLEU score of 36.1, which is 6.4 points higher than the original Transformer; In Test 2, the accuracy was 0.92 and the BLEU score was 35.4, which is better than the Transformer model. At different training scales, the proposed model has the lowest training time, with a validation stage inference time of 0.29 seconds, far superior to other models, and shows an accuracy of 0.89 in high ambiguity sentence recognition. The research results indicate that the model has significant advantages in flipped classroom language translation, and performs well in semantic understanding, generation efficiency, and robustness.

This study addresses the critical need to understand factors driving student loyalty in Physical Education (PE). Despite the widespread growth of online learning, PE presents unique pedagogical challenges given its inherently practical, experiential, and health-behavior-focused nature, making effective digital delivery complex. Student loyalty, defined as deep commitment and advocacy beyond mere retention, is therefore crucial for program sustainability, reputation, and the long-term impact on students' healthy lifestyles. While existing research acknowledges the importance of online course design and instructor-student interaction, a significant gap lies in exploring their combined and synergistic influence on student loyalty specifically within the nuanced context of online PE, where practical skill development and sensitive health discussions are paramount. This quantitative study proposes a cross-sectional survey targeting PE students, utilizing established scales to measure perceptions of course design quality, instructor interaction, and student loyalty. Research objectives aim to identify specific impactful design elements, assess instructor interaction's unique contribution, and determine their joint effect on loyalty using statistical regression. Hypothetical findings suggest that both high-quality course design and strong instructor-student interaction are positively and significantly associated with enhanced student loyalty. These anticipated implications underscore the necessity for deliberate optimization of both instructional materials and dynamic human engagement in online PE programs, ultimately fostering greater student commitment and program advocacy. Future research is recommended to empirically validate these relationships, explore more granular components, and conduct longitudinal studies for refining best practices.

This study explored the purported benefits of AI-powered personalized feedback systems on university students' motor skill development and self-efficacy within physical education (PE) in Heilongjiang, China. Employing a quantitative, quasi-experimental design, the research sought to compare an AI-feedback group against a control receiving traditional instruction. While the analysis reported robust, statistically significant improvements across all measured motor skill performance indicators and substantial gains in self-efficacy within the AI-feedback group, a critical interpretation is warranted. These preliminary findings, though seemingly positive, originate from a design that, by its quasi-experimental nature, may not fully eliminate confounding variables inherent to educational settings. The reported "very strong" and "significant" improvements in the experimental group, while statistically compelling, lack direct comparative between-group statistical measures in the provided summary, thus preventing a definitive claim about AI's superiority over traditional methods based on this excerpt alone. While the internal gains within the AI group are clear, the extent to which these surpass the improvements of a rigorously controlled traditional group remains to be fully demonstrated. Nonetheless, the observed magnitude of change strongly suggests a notable impact, implying that AI-powered feedback holds considerable potential to address the logistical challenges of individualized instruction in large PE classes, thereby fostering enhanced learning outcomes and bolstering self-efficacy. Future research employing more rigorous comparative analyses with actual data and incorporating qualitative insights is crucial to fully validate and contextualize these promising preliminary results.

New media short video production is flourishing but labour-intensive and low-quality content creation. Manual editing, captioning, and highlight extraction are high in time and low in personalization. This study uses novel Transformer architecture to automate major parts of video production, such as summarization, captioning, and thumbnail selection. The self-attention mechanism in Transformers can well model long-range dependencies among video frames and audio tracks. The model is trained on a huge dataset of short videos to perform tasks such as speech-to-text and scene segmentation. The efficiency in content creation is increased, and viewer engagement is promoted. The accuracy scores are used for text generation and view duration for assessing video performance. The suggested transformer models achieve 96% accuracy on content engagement and half the editing time compared to RNN-based systems. The method allows creators and platforms to create high-quality videos at scale.