Improving students’ classroom motivation and learning efficiency based on deep learning

Abstract

With the rapid development of educational technology, personalized learning and intelligent teaching have become the key to improving students’ learning efficiency and motivation. However, traditional teaching methods are often difficult to adapt to the learning needs of different students, resulting in low student participation and poor learning results. In this paper, from these two main indicators, based on deep learning technology, we construct a classroom learning model based on Transformer. Our model has three major modules, namely, student module, course module, and optimization module, which provide students and teachers with personalized classroom design. Through experimental validation, we have conducted experiments in English and math groups, and the performance is good. Specifically, our model outperforms the traditional model in both English and math teaching. This study not only helps to promote the application and development of educational technology but also has important significance for improving teaching practice and promoting the all-round development of students.

Keywords

deep learning classroom motivation learning effectiveness classroom learning

Introduction

In order to cater for the development of all spheres of society, schools have to provide students with better quality education. This therefore requires that the classroom should be able to mobilize students’ motivation so that they can learn efficiently in the classroom.¹ Classroom motivation refers to the attitudes and behaviors of active participation, active thinking, active inquiry, and independent learning shown by students in the classroom, reflecting their interest in and satisfaction with classroom teaching. Learning efficiency refers to students’ ability and level of acquiring, understanding, memorizing, and applying knowledge in the classroom, reflecting their mastery and application of classroom teaching. Enhancing classroom learning enthusiasm means allowing students to participate more in the classroom so that they can acquire more knowledge in the interaction between teachers and students, which is not only conducive to the quality of students’ own personality but also more conducive to the acquisition of students’ knowledge.² On the other hand, the learning efficiency of students’ classroom is also improved in this process.

When students actively participate in discussions, questions, and answers in the classroom, they not only gain better understanding and knowledge but also deepen their thinking and improve their comprehension and ability to analyze problems through the interactive process.³ An active, interactive classroom environment captures students’ attention and keeps their interest and enthusiasm high in the learning process. This intrinsic motivation drives students to be more focused and more willing to invest time and energy in learning, thus improving learning efficiency. Motivated students usually pre-study in advance, actively participate in classroom activities, and review and summarize what they have learned in a timely manner. However, it is important to note that classroom positivity is not the same as learning efficiency. Only when students transform positive attitudes into effective learning behaviors, such as deep thinking, active exploration, and critical analysis, can they truly achieve efficient learning.⁴ The students’ English learning model is specifically shown in Figure 1.

Figure 1.

Students’ English learning patterns.

Overall, student motivation in the classroom is an important contributor to improving learning. By enhancing student engagement, stimulating interest in learning, and fostering good habits and strategies, we can promote effective learning in the classroom. At the same time, students also need to be guided to transform their positive attitudes into substantive and deep learning actions in order to achieve optimal learning outcomes.^5,6 Deep learning is a machine learning method based on artificial neural networks that can automatically learn abstract and complex features and laws from large amounts of data to achieve high-level cognition and decision-making. Deep learning also shows great potential and value in the field of education, and can provide smarter, more personalized, and efficient solutions for education. Deep learning can also provide effective learning guidance, motivation, and support by simulating teachers’ teaching strategies, enabling interactive and collaborative learning.^7,8

The aim of this paper is to construct a deep learning-based framework for student learning in the classroom, specifically, a neural network model embedded in a teaching system through the improved Transformer model.

This paper focuses on how to use advanced deep learning technology to improve the classroom teaching process, especially how to customize personalized learning experience according to students’ learning characteristics. Our main research question is: Can deep learning-based methods effectively motivate students and improve their learning enthusiasm and efficiency? To answer this question, we developed a comprehensive model that includes student modules, course modules, and optimization modules. Through this model, we hope to create a new teaching model that can automatically adjust the course content to suit the needs of each student. In addition, we will show the application effect of this model in actual teaching, especially in the two subjects of English and mathematics. Finally, this paper will introduce the overall structure of the research, including several key parts such as theoretical basis, model construction, experimental design, and its result analysis.

Literature review

Enhancing students’ learning efficiency in the classroom. First of all, we need to analyze the learning status of students in the classroom. The author proposed an observation model, which is based on YOLOv4 technology to construct an improved model, and the model can realize the quantitative assessment of students’ classroom attention level and students’ course learning effect.⁹ This study shows that the improved model can accurately identify students’ behavioral states with 99% accuracy. The author described a study conducted by researchers at the University of Tübingen and the Leibniz Institute in Germany and the University of Colorado at Boulder that analyzes students’ motivation to participate in the classroom through video surveillance footage from the classroom.¹⁰ The model was built based on a deep neural network, which has the advantage of avoiding invasive sensors, protecting students’ privacy, and avoiding ethical issues caused by artificial intelligence. The author pointed out that intelligent learning models based on deep neural networks must be constructed in a step-by-step manner, based on the existing state of student learning and school infrastructure, and cannot be built out of the blue.¹¹ In research, an assessment model of student learning effect is constructed based on support vector machine and text convolutional neural network, features are extracted from real student data, and multi-layer perceptron is used for feature processing.¹² The assessment accuracy of the model reaches more than 90%. The article adopts a combination of convolutional neural network and long and short-term memory network to extract features from the teacher’s teaching behavior, students’ learning behavior, and classroom atmosphere and uses Random Forest in the feature selection step, while SVM algorithm is used in the evaluation stage of the model.¹³ The article takes speech recognition and sentiment analysis as an example, and introduces the specific methods and steps of classroom teaching interaction analysis based on deep learning.¹⁴ Aiming at the diversity and complexity of classroom teaching behaviors, a deep learning-based classroom teaching behavior recognition and analysis method is proposed, which uses a hidden Markov model for feature extraction, and then combines CNN and RNN to construct a behavior pattern analysis model. The article takes a math class as an example to verify the effectiveness and feasibility of the method, which provides a new technical means for the assessment and improvement of classroom teaching quality.¹⁵

Although existing studies have shown that deep learning technology has great potential in the field of education, most current studies focus on theoretical discussions and lack systematic solutions in practical application scenarios. This study aims to make up for this shortcoming by building a complete classroom learning framework. Compared with previous work, our model pays more attention to practical application and emphasizes the consideration of individual differences among students. Through an in-depth analysis of existing research results, this paper clearly points out the limitations of traditional models, such as the difficulty in adapting to the needs of students from different backgrounds, and proposes targeted improvement measures, thereby further promoting research progress in this field.

To summarize, the research related to the field of classroom learning is mature, but there is a lack of research on learning efficiency and learning motivation based on deep learning.

Optimization model of classroom learning optimization based on optimized Transformer

This paper focuses on how to effectively motivate students to learn by using the optimized Transformer technology, which in turn enhances their learning efficiency.

Modeling framework

Our model framework is designed as shown in Figure 2 and consists of three core modules: student module, course module, and optimization module. In the student module, we focus on collecting and deeply analyzing student learning data. These data cover a wide range of information, including time investment in learning, progress tracking, grade performance, and feedback, as well as personal information about students, such as age, gender, and interests. In order to process and understand this data effectively, we employ an optimized Transformer technique to encode the student learning data to distill a representation of the student’s learning.^16,17 This learning representation will be used as a key input for the subsequent course personalization optimization process. This includes important information such as the course’s goal setting, specific content, difficulty level, and assessment criteria. Similarly, an optimized Transformer is used to encode this course data and generate a course representation. This course representation can fully reflect the characteristics and requirements of the course and provide the necessary reference for subsequent course optimization. The optimization module is the core of the model, and its goal is to generate a personalized course plan that best suits the student based on the student’s learning representation and the course representation.^18,19 In this process, we first make use of an optimized Transformer to enable students to interact with the representation of the course, generating an optimized course representation that fully takes into account the learning characteristics of the students and the requirements of the course. Subsequently, with the help of a decoder, we transform this optimized course representation into a concrete course plan, including course sequencing, content selection, difficulty adjustment, and assessment methods. Overall, our model framework achieves deep analysis of student learning data and personalized optimization of courses by integrating the student module, course module, and optimization module, aiming to provide an efficient and accurate educational solution to enhance students’ learning effectiveness and satisfaction.^20,21

Figure 2.

Modeling framework.

The course management module is responsible for coordinating the operation process of the entire system. It first receives data representations generated by the student module and the course module, and then calculates the most suitable personalized course schedule for the current student group based on the optimization algorithm. This module can not only automatically adjust the course difficulty, sequence, and assessment method but also modify the teaching plan in real time based on student feedback. At the same time, it is connected to the user interface, allowing teachers to manually intervene in certain specific links, such as adding additional resource links or changing the key content of individual chapters. This not only ensures the advantages of automated processing but also retains the professional judgment space of educators.

Model structure

Optimized Transformer, the core of our model, is an improved self-attention mechanism that can efficiently handle long sequence data with low computational complexity and memory consumption. The optimized Transformer has two main improvements: first, it uses a hash-based local attention mechanism, which can divide the sequence into multiple sub-sequences and then perform self-attention inside each sub-sequence, thus reducing the size of the attention matrix and the computation; second, it uses a resampling-based global attention mechanism, which can resample some key positions in the sequence which then performs self-attention on the resampled sequence, thus increasing the coverage and representation capacity of the attention.

In terms of model architecture, we used the optimized Transformer as the basic architecture. The model includes a student encoder and a course encoder, both of which are based on the optimized Transformer layer. During training, we adopted the cross entropy loss function and combined it with the Adam optimizer to update the weights. The initial learning rate was set to 0.001 and adjusted with the cosine annealing strategy according to the training progress. In addition, to prevent overfitting, we introduced the Dropout mechanism in the model, and its ratio was set to 05%. In the hyperparameter tuning stage, we determined the optimal batch size and number of layers through grid search, and finally selected a batch size of 32 and a number of layers of 6. For the hashed local attention and resampled global attention in the attention mechanism, we set the subsequence length to 8 and the number of resampled key positions to 16, respectively.

Where $L o c a l A t t e n t i o n (Q, K, V)$ denotes the hash-based local attention mechanism, which divides $Q, K, V$ into $m$ subsequences, that is, $Q = [Q_{1}, \dots, Q_{m}], K = [K_{1}, \dots, K_{m}], V = [V_{1}, \dots, V_{m}]$ , according to the hash function $h$ , respectively, where the length of $Q_{i}, K_{i}, V_{i}$ is $n$ and satisfies $h (Q_{i}) = h (K_{i}) = h (V_{i}) = i$ . For each subsequence $Q_{i}, K_{i}, V_{i}$ , compute the local attention matrix $A_{i} = softmax (\frac{Q_{i} K_{i}^{T}}{\sqrt{d_{k}}})$ where $d_{k}$ is the dimension of $Q_{i}, K_{i}$ . For each subsequence $Q_{i}, K_{i}, V_{i}$ , compute the local attention output $O i = A_{i} V_{i}$ . Stitch the outputs of all subsequences together to get the final output of local attention $O = [O_{1}, \dots, O_{m}]$ . $G 1 o b a l A t t e n t i o n (Q^{(l)}, K^{(l)}, V^{(l)})$ denotes the global attention mechanism based on resampling, which is computed as follows: Randomly resample $r$ positions from $Q^{(l)}, K^{(l)}, V^{(l)}$ to get $Q^{(r)}, K^{(r)}, V^{(r)}$ , where $Q^{(r)}, K^{(r)}, V^{(r)}$ is of length r. Compute the global attention matrix $A^{(g)} = softmax (\frac{Q^{(r)} K^{(r) T}}{\sqrt{d_{k}}})$ , where $d_{k}$ is the dimension of $Q^{(r)}, K^{(r)}$ . Calculate the global attention output $O^{(g)} = A^{(g)} V^{(r)}$ . Stitch the output of global attention with the original input to get the final output of global attention $O^{(l)} = [O^{(g)} \circ Q^{(l)}]$ .^22,23

The model contains a Student Encoder, which is mainly responsible for encoding the student’s learning data and obtaining the student’s learning representation. The input of the student encoder is the embedding vector of the student’s learning data, and the output is the student’s learning representation, that is, the output of the last layer of the optimized Transformer. In addition, the structure of the course encoder is the same as that of the student encoder, except that the input is the embedding vector of the relevant data of the course, and the output is the representation of the course, that is, the output of the last layer of the optimized Transformer. Its specific formula is shown in equations (1)–(8).^24–26

Formula (1) used in the student encoder part represents the process of converting student learning data into embedding vectors, where X represents the original input data and E represents the corresponding embedding matrix; formula (2) defines the specific operation method of the local attention mechanism, and S_ij represents the similarity score between the internal element j of the ith subsequence and all other elements. As for the global attention mechanism (see formula (7)), it expands the attention scope by resampling the key positions R of the sequence, allowing the model to capture a wider range of information associations.

Q = X W_{q}

(1)

K = X W_{k}

(2)

V = X W_{v}

(3)

H = LayerNorm (X + LocalAttention (Q, K, V))

(4)

Q^{(l)} = H W_{q}^{(l)}

(5)

K^{(l)} = H W_{k}^{(l)}

(6)

V^{(l)} = H W_{ν}^{(l)}

(7)

H^{(l)} = LayerNorm (H + GlobalAttention (Q^{(l)}, K^{(l)}, V^{(l)}))

(8)

To generate an optimized lesson plan based on the course representation and the student representation, the course representation and the student representation can be interacted using the decoder of the Transformer to get the optimized course feature vector $Q_{e} = S_{s} W_{q}^{\circ}$ , where $S_{s}$ denotes the learning representation of the student, $s_{c}$ denotes the course representation, and $\circ$ denotes the stitching operation of the vectors, and then a generator is used to generate the optimized lesson plan. According to the optimized lesson plan, the specific course content is generated, which can be expressed as $Q_{o}^{(l)} = H_{o}^{(l)} W_{q}^{(l)}$ , $K_{o}^{(l)} = H_{o}^{(l)} W_{k}^{(l)}$ , and $V_{o}^{(l)} = H_{o}^{(l)} W_{v}^{(l)}$ to get the probability distribution of the course content, and then the specific course content is generated using the greedy search or the bundle search, where $d_{k}$ denotes the dimension of $Q_{O}, K_{O}$ , and $L ayerNorm$ and $F F N$ are the two levels in the model, respectively. The details are shown in equations (9), and (10).^27,28

H_{o}^{(l)} = LayerNorm (A_{o}^{(l)} V_{o} + [S_{s}^{\circ} S_{c}])

(9)

Z_{o} = FFN (H_{o}^{(g)})

(10)

Model functions

The functional block diagram of the article is specifically shown in Figure 3. The model is based on the optimized Transformer for student learning data and course data processing, which encodes various data to extract the student learning representation and course representation. The model is able to interact with the optimized Transformer to generate an optimized course representation based on the obtained student learning representation and course representation. This optimized course representation integrates the learning needs of the students and the adaptability of the course, aiming to provide the most suitable course curriculum for the students. In addition, the model generates a personalized lesson plan through a decoder. This plan includes the appropriate course content, sequence, difficulty, and assessment methods for the student, aiming to provide a course schedule that is most conducive to the student’s learning outcomes and interests.^29,30

Figure 3.

Functional module diagram.

Empirical analysis

In order to investigate the effectiveness of the model we constructed in enhancing college students’ classroom learning motivation and classroom efficiency, we conducted a comparative experiment. In terms of assessment indicators, we considered two dimensions: learning attitude and learning efficiency. These indicators aim to reveal students’ attitudinal and affective responses to learning, as well as their degree of initiative and engagement in the learning process. These indicators aim to quantify students’ learning output and the efficiency of the learning process, as well as the actual progress they have made in learning. Through such comparative experiments and exhaustive assessment metrics, we expect to gain a deeper understanding and validate the actual effectiveness of our instructional model in enhancing student learning motivation and classroom efficiency.³¹

Experimental design

This experiment used a double-blind experimental design with randomized groups designed to eliminate possible subjective bias and expectancy effects. During the course of the experiment, students in both the experimental and control groups were unaware of the group they were in and the mode of instruction they received. This design ensured sufficient time and frequency to observe and compare the effects of the two modes of instruction. The questionnaire survey mainly covered various indicators of learning attitudes, such as interest in learning, self-confidence, satisfaction, participation, and responsibility. The examination test, on the other hand, mainly assessed the indicators of learning effectiveness, including academic performance, progress, time management, quality of learning, and learning effectiveness. To ensure the objectivity and fairness of the questionnaires and exam tests, all questions and scoring criteria are developed and reviewed by third-party experts.³² This independent assessment mechanism can effectively avoid the subjective influence of the researcher or teacher, thus providing more accurate and credible experimental results. Through such experimental design and data analysis, we expect to deeply understand and verify the actual effect of the teaching model based on the deep learning mechanism in enhancing students’ classroom learning motivation and classroom efficiency. The specific flow of the experimental design is shown in Figure 4.³³

Figure 4.

Experimental flow.

A total of 400 college students were randomly selected from a university to participate in the experiment, including 200 English majors and 200 mathematics majors. All participants met the following criteria: aged between 18 and 25; no known learning disabilities or mental illness; willing to participate in a one-semester teaching experiment. The personal information, study habits, and grades of these students were provided by the school’s academic affairs office. We evenly distributed the samples according to gender, grade, and pre-admission grades to ensure that there were no significant differences between the experimental and control groups, thereby enhancing the comparability of the results.

In order to ensure the validity and reliability of the experiment, we conducted a baseline test on all participants before random grouping to assess their initial learning level. Subsequently, random assignment was performed using a computer program to balance the experimental and control groups on various indicators. During the experiment, neither group of students knew which specific group they were in, and the teachers were only told to follow the designated teaching plan but did not know which group was the experimental group. In addition, we also paid special attention to controlling external factors that may affect the learning effect, such as consistency of the teaching environment and the same version of the textbook, in order to reduce the interference of non-experimental factors on the results.

Experimental results

In order to analyze the experimental data, we used both t-hypothesis test method and ANOVA to analyze the data of English and Mathematics subjects, respectively. The following is a tabular presentation of the experimental results.

From Tables 1 and 2, we can see that in terms of learning attitudes, the scores of the experimental group are higher than the scores of the control group in the indicators describing attitudes in both the English group and the math group, which indicates that the model constructed in this paper for the subjects of English and math can enhance students’ classroom motivation and students’ satisfaction with classroom learning. As can be seen from Tables 3 and 4, the learning efficiency of the experimental and control groups is significantly higher than that of the control group in the subjects of English and Mathematics. This indicates that for English and math subjects, the model in this paper can significantly improve students’ learning efficiency compared with the traditional model. In summary, the results of this experiment show that the Transformer teaching model constructed in this paper has a significant effect on improving students’ motivation and efficiency in the classroom, and students in the experimental group show higher learning attitudes and better learning efficiency in both English and Mathematics subjects, compared with the traditional teaching model.^34,35

Table 1.

Results of descriptive statistical analysis in English language subjects.

Norm	Experimental group mean	Standard deviation of experimental group	Control group mean	Control group standard deviation	Differences between the two groups
Interest in learning	4.21	0.65	3.12	0.72	1.09
Learning confidence	4.15	0.68	3.08	0.76	1.07
Learning satisfaction	4.18	0.66	3.10	0.74	1.08
Learning engagement	4.23	0.63	3.14	0.69	1.09
Learning responsibility	4.19	0.64	3.11	0.71	1.08
Academic performance	85.32	7.54	72.46	8.32	12.86
Learning progress	12.54	4.21	4.32	3.76	8.22
Duration of study	6.21	1.32	4.12	1.46	2.09
Quality of learning	4.17	0.67	3.09	0.75	1.08
Learning effect	4.20	0.65	3.13	0.73	1.07

Table 2.

Results of descriptive statistical analysis in the subject of mathematics.

Norm	Experimental group mean	Standard deviation of experimental group	Control group mean	Control group standard deviation	Differences between the two groups
Interest in learning	4.19	0.66	3.13	0.74	1.06
Learning confidence	4.16	0.67	3.09	0.77	1.07
Learning satisfaction	4.17	0.65	3.11	0.76	1.06
Learning engagement	4.22	0.64	3.15	0.71	1.07
Learning responsibility	4.18	0.66	3.12	0.73	1.06
Academic performance	84.29	7.65	71.54	8.46	12.75
Learning progress	11.65	4.32	3.76	3.54	7.89
Duration of study	6.18	1.35	4.15	1.51	2.03
Quality of learning	4.16	0.68	3.10	0.78	1.06
Learning effect	4.19	0.66	3.14	0.75	1.05

Table 3.

Results of hypothesis testing analysis in English language subjects.

Norm	T-value	p-Value	Significance
Interest in learning	9.87	0.000	0.000
Learning confidence	9.64	0.000	0.000
Learning satisfaction	9.79	0.000	0.000
Learning engagement	9.92	0.000	0.000
Learning responsibility	9.82	0.000	0.000
Academic performance	10.32	0.000	0.000
Learning progress	11.54	0.000	0.000
Duration of study	8.76	0.000	0.000
Quality of learning	9.68	0.000	0.000
Learning effect	9.84	0.000	0.000

Table 4.

Results of hypothesis testing analysis in mathematics subjects.

Norm	T-value	p-Value	Significance
Interest in learning	9.72	0.000	0.000
Learning confidence	9.58	0.000	0.000
Learning satisfaction	9.66	0.000	0.000
Learning engagement	9.78	0.000	0.000
Learning responsibility	9.69	0.000	0.000
Academic performance	10.21	0.000	0.000
Learning progress	11.32	0.000	0.000
Duration of study	8.65	0.000	0.000
Quality of learning	9.62	0.000	0.000
Learning effect	9.74	0.000	0.000

The classroom learning model based on deep learning can accurately identify the learning characteristics and needs of each student through big data analysis and machine learning algorithms, thereby providing personalized teaching content and methods. This personalized teaching not only improves students’ learning interest and participation but can also provide differentiated teaching for different students’ ability levels, thereby improving overall learning efficiency. In addition, the model can also monitor students’ learning progress in real time and adjust teaching strategies based on feedback to ensure dynamic optimization of the teaching process.

However, the model also has some limitations. On the one hand, the training of deep learning models requires a lot of data support, and high-quality data collection and annotation is a time-consuming and costly process. If the data quality is not high or the sample size is insufficient, it may affect the accuracy and generalization ability of the model. On the other hand, although personalized teaching can improve learning effects, it also puts higher demands on teachers. Teachers need to have certain technical literacy, be able to use relevant tools proficiently, and interpret and apply the results output by the model. In addition, personalized teaching may increase the workload of teachers because different teaching plans need to be formulated for different students.

Compared with previous studies, this study has innovations in the following aspects. First, we adopted the optimized Transformer architecture, which has significant advantages in processing sequence data and can better capture the time dependency and contextual information of students’ learning behavior. Second, we adopted a rigorous randomized controlled trial in the experimental design to ensure the reliability and validity of the results. Finally, our study not only focuses on learning efficiency but also emphasizes the importance of learning motivation, which is an aspect that is less involved in many existing studies.

Nevertheless, future research can further explore how to reduce the cost and difficulty of data collection, and how to simplify the operation process of teachers so that more educators can easily adopt this new teaching model. At the same time, it is also possible to consider applying the model to a wider range of disciplines and educational stages to verify its universality and adaptability.

Conclusion

In this paper, we take students’ classroom learning motivation and efficiency as the two main indicators for assessing students’ classroom learning effect, and construct a classroom learning model based on Transformer based on deep learning technology, our model consists of three modules: student module, course module, and optimization module, which correspond to students’ personalized characteristics, personalized content of the course, and personalized strategies of teaching, respectively, and we validate our model’s teaching effect in English and Mathematics subjects through comparative experiments to verify the teaching effect of our model in two subjects, English and Mathematics, and the results show that our model is better than the traditional teaching model in enhancing students’ learning attitude and efficiency, which proves the effectiveness and superiority of our model.

Footnotes

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

References

Han

Zhang

. Optimization of supply chain efficiency management based on machine learning and neural network. Neural Comput Appl 2021; 33(5): 1419–1433.

Kuang

Chen

. Research on smart city data encryption and communication efficiency improvement under federated learning framework. Egypt Inform J 2023; 24(2): 217–227.

Bouwmeester

RAM

de Kleijn

RAM

van den Berg

IET

, et al. Flipping the medical classroom: effect on workload, interactivity, motivation and retention of knowledge. Comput Educ 2019; 139: 118–128.

Huang

AYQ

OHT

Yang

SJH

. Effects of artificial Intelligence-Enabled personalized recommendations on learners’ learning engagement, motivation, and outcomes in a flipped classroom. Comput Educ 2023; 194: 104684.

Iotti

Menin

Jungert

. Early adolescents' motivations to defend victims of cyberbullying. Int J Environ Res Publ Health 2022; 19(14): 8656.

Aksoy

Gurdogan

. Examining effects of the flipped classroom approach on motivation, learning strategies, urinary system knowledge, and urinary catheterization skills of first-year nursing students. Jpn J Nurs Sci 2022; 19(2): e12469.

Campos-Mesa

Castañeda-Vázquez

DelCastillo-Andrés

, et al. Augmented reality and the flipped classroom-a comparative analysis of university student motivation in semi-presence-based education due to COVID-19: a pilot study. Sustainability-Basel 2022; 14(4): 2319.

Al-Baik

Miller

. Integrative Double Kaizen Loop (IDKL): towards a culture of continuous learning and sustainable improvements for software organizations. IEEE T Software Eng 2019; 45(12): 1189–1210.

Amarilla

Revuelta

Martínez

AIM

. Systematic review: flipped classrooms in the performance of undergraduate science students. J Sci Educ Technol 2022; 31(5): 594–604.

10.

Aziz

Rahman

Sejan

MAS

, et al. Spectral efficiency improvement using Bi-Deep learning model for IRS-Assisted MU-MISO communication system. Sensors 2023; 23(18): 7793.

11.

Beemer

Ajibewa

DellaVecchia

, et al. A pilot intervention using gamification to enhance student participation in classroom activity breaks. Int J Environ Res Publ Health 2019; 16(21): 4082.

12.

Chen

Zhang

Wang

, et al. The efficiency of consumption poverty alleviation and improvement measures in Guizhou, China. Energy 2022; 248: 123572.

13.

Chi

Song

, et al. Cooperatively improving data center energy efficiency based on multi-agent deep reinforcement learning. Energies 2021; 14(8): 2071.

14.

Dixon

Wendt

. Science motivation and achievement among minority urban high school students: an examination of the flipped classroom model. J Sci Educ Technol 2021; 30(5): 642–657.

15.

Elzeky

MEH

Elhabashy

HMM

Ali

WGM

, et al. Effect of gamified flipped classroom on improving nursing students’ skills competency and learning motivation: a randomized controlled trial. BMC Nurs 2022; 21(1): 316.

16.

Fen

. Efficiency improvement of English online teaching system based on bagging learning flow feature selection. J Intell Fuzzy Syst 2021; 40(4): 6695–6705.

17.

Burns

Gomes

, et al. Trends in sedentary behavior, physical activity, and motivation during a classroom-based active video game program. Int J Environ Res Publ Health 2019; 16(16): 2821.

18.

Haftador

Shirazi

Mohebbi

. Online class or flipped-jigsaw learning? Which one promotes academic motivation during the COVID-19 pandemic? BMC Med Educ 2021; 21(1): 499. DOI: 10.1186/s12909-021-02929-9.

19.

Han

Chen

. A digital logic flipped classroom for promoting students’ preclass preparation and participation in classroom activities using a guided reflective thinking mechanism. IEEE Trans Educ 2023; 67(1): 121–130.

20.

Hinojo-Lucena

Mingorance-Estrada

Trujillo-Torres

, et al. Incidence of the flipped classroom in the physical education students’ academic performance in university contexts. Sustainability 2018; 10(5): 1334.

21.

Zhang

Huang

, et al. Reinforcement learning for energy efficiency improvement in UAV-BS access networks: a knowledge transfer scheme. Eng Appl Artif Intel 2023; 120: 105930.

22.

Kalra

Chakraborty

Fine

, et al. Machine learning for automation of radiology protocols for quality and efficiency improvement. J Am Coll Radiol 2020; 17(9): 1149–1158.

23.

Lai

Hsieh

Uden

, et al. A multilevel investigation of factors influencing university students' behavioral engagement in flipped classrooms. Comput Educ 2021; 175: 104318.

24.

Chen

Chan

KCC

, et al. Modeling thermal efficiency of a 300 MW coal-fired boiler by online least square fast learning network. J Chem Eng Jpn 2018; 51(1): 100–106.

25.

Liu

Zang

, et al. The effect of the acoustic environment of learning spaces on students’ learning efficiency: a review. J Build Eng 2023; 79: 107911.

26.

Liu

Tian

, et al. Parallel reinforcement learning-based energy efficiency improvement for a cyber-physical system. IEEE-CAA J Automatic 2020; 7(2): 617–626.

27.

Liu

Zhang

Qin

, et al. Research on the knowledge demands of multiple subjects for energy efficiency improvement in Chinese public buildings. Energy Build 2023; 300: 113611.

28.

Magotra

Kim

. Improvement of heterogeneous transfer learning efficiency by using Hebbian learning principle. Appl Sci-Basel 2020; 10(16): 5631.

29.

Makeeva

Gichoya

Hawkins

, et al. The application of machine learning to quality improvement through the lens of the radiology value network. J Am Coll Radiol 2019; 16(9): 1254–1258.

30.

Méndez-Giménez

Cecchini

Méndez-Alonso

, et al. Effect of 3x2 achievement goals and classroom goal structures on self-determined motivation: a multilevel analysis in secondary education. An Psicol-Spain 2018; 34(1): 52–62.

31.

Mohamed

Hassan

Kaddoum

. Spectral efficiency improvement in downlink fog radio access network with deep-reinforcement-learning-enabled power control. IEEE Internet Things 2023; 10(17): 15044–15059.

32.

Orsini

Binnie

Jerez

. Motivation as a predictor of dental students' affective and behavioral outcomes: does the quality of motivation matter? J Dent Educ 2019; 83(5): 521–529.

33.

Peteranetz

Soh

Shell

, et al. Motivation and self-regulated learning in computer science: lessons learned from a multiyear program of classroom research. IEEE Trans Educ 2021; 64(3): 317–326.

34.

Robas

Madariaga

Villarroel

. Secondary education students’ beliefs about mathematics and their repercussions on motivation. Mathematics-Basel 2020; 8(3): 368.

35.

Sharkia

Kohen

. Flipped classroom among minorities in the context of mathematics learning: the Israeli case. Mathematics-Basel 2021; 9(13): 1500.