This study explores the spatial distribution pattern of educational infrastructure development across districts and cities in North Sumatra, identifying significant disparities between urban and rural areas. The study aims to: (1) determine the distribution of educational development across districts and cities, (2) analyze global spatial autocorrelation, and (3) identify priority locations for educational development policies in North Sumatra Province. The methodology includes quantile analysis, Moran’s Global Index, and Local Indicators of Spatial Autocorrelation (LISA) using GeoDa software to address spatial autocorrelation. The results indicate that there are nine areas with a low School Participation Rate Index (SPRI), eleven areas with a low School Facilities and Infrastructure Index (SFII), and eleven areas with a low Regional Education Index (REI). Spatial autocorrelation analysis reveals that SFII shows positive spatial autocorrelation, while SPRI and REI exhibit negative spatial autocorrelation, indicating a high level of inequality between regions. Labuhan Batu Selatan and Labuhan Batu are identified as priorities for the provincial government in overseeing educational development policies.

Language is fundamental to human communication, allowing individuals to express and exchange ideas, thoughts, and emotions. In early childhood, some children experience communication disorders that impede their ability to articulate words correctly, posing significant challenges to their learning and development. This issue is exacerbated in developing countries, where limited resources and a lack of technological tools hinder access to effective speech therapy. Traditional speech therapy remains vital, but the latest technological advancements have introduced robotic assistants to enhance therapy for communication disorders. Despite their potential, these technologies are often inaccessible in developing regions due to high production costs and a lack of sustainable manufacturing models. For these reasons, this paper presents “FONA,” a robotic assistant that employs rule-based expert systems to provide tactile, auditory, and visual stimuli. FONA supports children aged 3 to 6 in speech therapy by delivering exercises such as syllable production, word formation, and pictographic storytelling of various phonemes. Notably, FONA was successfully tested on children with cochlear implants, reducing the number of sessions required to produce isolated phonemes. The paper also introduces an innovative analysis of the Make To Order (MTO) manufacturing system for producing FONA in developing countries. This analysis explores two key perspectives: collaborative networks and entrepreneurship, offering a sustainable production model. In a pilot experiment, FONA significantly improved children’s attention spans, increasing the period by 17 min. Furthermore, the economic analysis demonstrates that producing FONA through collaborative networks can significantly reduce costs, making it more accessible to institutions in developing countries. The findings suggest that the project is viable for a five-year period, providing a sustainable and effective solution for addressing communication disorders in children.

With the continuous promotion and deepening of quality education, new teaching goals have been proposed for major universities and teachers, requiring teachers not to blindly pursue the academic performance of college students as the goal, but to achieve the comprehensive development of college students as the main teaching goal. Therefore, teachers need to actively transform educational concepts, transform educational methods, enrich classroom content, and provide high-quality teaching classrooms for college students, Help college students improve in all aspects. For college students, it is not only necessary to cultivate correct worldviews and values, establish positive life goals and attitudes, but also to enhance their resistance to pressure when facing society. Therefore, when teaching, teachers not only need to explain knowledge, but also serve as guides on the life path of college students, helping them guide and improve their ideological and moral character, Thus achieving significant development of ideological and political education in universities.

This study analyzes the perception of university students regarding the use of virtual reality (VR) in higher education, focusing on their level of knowledge, usage, perceived advantages and disadvantages, as well as their willingness to use this technology in the future. Using a mixed-methods approach that combines questionnaires and semi-structured interviews, both quantitative and qualitative data were collected to provide a comprehensive view of the subject. The results indicate that while students have a basic understanding of VR, its use in the educational context is limited. A considerable number of students recognize VR’s potential to enhance the learning experience, particularly in terms of immersion and engagement. However, significant barriers to adoption were identified, such as technical issues, the high cost of equipment, and inadequate access to technological infrastructure. Additionally, there is a need for broader training for both students and faculty to ensure the effective use of this technology in academic environments. The semi-structured interviews confirmed that perceptions of VR vary depending on prior exposure to the technology and access to resources. Despite the challenges, most students appreciate VR’s potential to enrich learning, although its effective adoption will depend on overcoming the identified barriers. The study concludes that strategies must be implemented to facilitate the integration of VR into higher education, thus optimizing its impact on the teaching-learning process.

With the in-depth development and widespread application of educational informatization, digital education has also become one of the important features of educational modernization. Designing and completing a visual teaching system based on Web technology is of great significance for promoting further reform and development of teaching, especially for achieving remote education, which has great application value. Based on visual teaching needs analysis and B/S architecture, effective system development is achieved through Access database. According to the specific needs of teaching functions, the system can be divided into multiple modules, and the management and login of teaching resources for users can also be smoothly achieved. This has important research value for achieving the goal of remote visualization of teaching.

This study conducts a comparative analysis of various machine learning and deep learning models for predicting order quantities in supply chain tiers. The models employed include XGBoost, Random Forest, CNN-BiLSTM, Linear Regression, Support Vector Regression (SVR), K-Nearest Neighbors (KNN), Multi-Layer Perceptron (MLP), Recurrent Neural Network (RNN), Bidirectional LSTM (BiLSTM), Bidirectional GRU (BiGRU), Conv1D-BiLSTM, Attention-LSTM, Transformer, and LSTM-CNN hybrid models. Experimental results show that the XGBoost, Random Forest, CNN-BiLSTM, and MLP models exhibit superior predictive performance. In particular, the XGBoost model demonstrates the best results across all performance metrics, attributed to its effective learning of complex data patterns and variable interactions. Although the KNN model also shows perfect predictions with zero error values, this indicates a need for further review of data processing procedures or model validation methods. Conversely, the BiLSTM, BiGRU, and Transformer models exhibit relatively lower performance. Models with moderate performance include Linear Regression, RNN, Conv1D-BiLSTM, Attention-LSTM, and the LSTM-CNN hybrid model, all displaying relatively higher errors and lower coefficients of determination (R²). As a result, tree-based models (XGBoost, Random Forest) and certain deep learning models like CNN-BiLSTM are found to be effective for predicting order quantities in supply chain tiers. In contrast, RNN-based models (BiLSTM, BiGRU) and the Transformer show relatively lower predictive power. Based on these results, we suggest that tree-based models and CNN-based deep learning models should be prioritized when selecting predictive models in practical applications.