key: cord-0473326-dkshij8d
authors: Elnagar, Ashraf; Afyouni, Imad; Shahin, Ismail; Nassif, Ali Bou; Salloum, Said A.
title: The empirical study of e-learning post-acceptance after the spread of COVID-19: A multi-analytical approach based hybrid SEM-ANN
date: 2021-11-27
journal: nan
DOI: nan
sha: 5286940da72dff0089bdf2639a6dae52265b14b0
doc_id: 473326
cord_uid: dkshij8d

There are several reasons why the fear of vaccination has caused population rejection. Questions have been raised by students regarding the effectiveness of vaccines, which in turn has led to vaccination hesitancy. Students perceptions are influenced by vaccination hesitancy, which affects the acceptance of e-learning platforms. Hence, this research aimed to examine the post-acceptance of e-learning platforms on the basis of a conceptual model that employs different variables. Distinct contribution is made by every variable to the post-acceptance of e-learning platforms. A hybrid model was used in the current study in which technology acceptance model (TAM) determinants were employed along with other external factors such as fear of vaccination, perceived routine use, perceived enjoyment, perceived critical mass, and self-efficiency which are directly linked to post-acceptance of e-learning platforms. The focus of earlier studies on this topic has been on the significance of e-learning acceptance in various environments and countries. However, in this study, the newly-spread use of e-learning platforms in the gulf area was examined using a hybrid conceptual model. The empirical studies carried out in the past mainly used structural equation modelling (SEM) analysis; however, this study used an evolving hybrid analysis approach, in which SEM and the artificial neural network (ANN) that are based on deep learning were employed. The importance-performance map analysis (IPMA) was also used in this study to determine the significance and performance of each factor. The proposed model is backed by the findings of data analysis.

Various challenges were experienced by higher education institutions during the COVID-19 pandemic, even after the availability of vaccines. Due to these challenges, a major change was required in the field of teaching and learning. During the pandemic, majority of the schools and universities shifted to virtual classrooms from the traditional physical classroom, which was the only way teaching methods and goals could be applied [1] - [3] . University students shifted to an entirely different setting, where they experienced a number of issues [4] , [5] . When the challenges experienced by students are comprehended, the method that can be used to evaluate the students' understanding, success and accomplishment during the second wave of the coronavirus pandemic can be established [6] .

It should be noted that this challenge continues to be experienced even after the vaccination has become available. This arises from the fact that the population has rejected the vaccinatio n.

Vaccination hesitancy refers to the hesitation shown by people towards different vaccines in several countries worldwide. It is a critical factor that impacts the world in general and the education system in specific. It suggests that a delay is experienced in the acceptance or rejection of vaccines, even though vaccines have become available. This view differs across certain locations and over time and is closely linked to confidence and convenience. It is also influe nced by contextual factors, group impact, individual beliefs, and people's faith in health services.

Vaccination hesitancy differs from one specific user to another, based on their self-perceptio n, perceived risk, and workplace. Therefore, vaccination hesitancy is the key issue being experienced regarding the COVID-19 vaccine uptake, which means that several hurdles will be experienced by the vaccine against COVID-19 in a post-crisis scenario [7] - [9] . Opposing vaccination hesitancy is vaccination confidence, which refers to something that can be accomplished "in itself".

Confidence in vaccines relies on the trust in the healthcare system as well as trust in a sociopolitical context. Users may not have a consistent confidence in vaccination because of the perceived risks associated with immunisation. Rather, it may give rise to smaller vaccinatio n coverage and loss of immunity [10] - [12] .

On the basis of the aforementioned assumptions, the objective of this research was to examine how vaccination hesitancy or fear affects post-acceptance of e-learning platforms, where the challenge continues to be critical and obvious because of vaccination rejection. A conceptual framework was formulated with the goal of attaining the study objective, which caters to the two factors being discussed, "i.e. post-acceptance and fear of vaccination". A few external variables were included in the conceptual model, because they have a direct relationship with post-acceptance, namely "perceived enjoyment, perceived routine use, self-efficiency and fear of vaccinations" [13] - [16] .

Consequently, the precise contribution of the present study can be summed up as follows: the study first examined the impact of vaccination fear or hesitancy in the educational sector. For this, an integrated research model was used; a conceptual model was formulated in which the TAM acceptance model [17] was integrated with the flow theory [18] , [19] to show the importance and predictability of the findings. Second, in contrast to the earlier empirical studies that mainly depended on structural equation modeling (SEM) analysis, an emerging hybrid analysis approach was used in this method that employs SEM and artificial neural network (ANN) based on deep learning. The importance-performance map analysis (IPMA) was also used in this research to determine the significance and performance of every factor. Third, vaccination fear and hesitancy is a critical factor and has different effects based on aspects such as the age and gender of people and the financial status of a country. A recent study revealed that there is greater vaccine hesitancy in low-income countries and among young females and older adults [20] . Lastly, the model was extended in this study to consider external variables that have a close relationship with the postacceptance stage [20] , which include critical mass and daily routine. We believe that this is the first study that seeks to examine the post-acceptance of e-learning platforms based on an integrated model in which fear of vaccination is the key variable, with the aim of filling a major research gap in the relevant literature.

According to the literature review, the impact of COVID-19 on the various educational e-learning platforms have been analysed in earlier studies. These platforms included Moodle, Zoom Microsoft Teams, Google Classroom, virtual reality applications, and so on. The studies showed that all these platforms were effective while the pandemic was spreading and offered a viable solution to the issue [21] - [23] . In most of the previous studies, TAM has been the most influentia l model used. The focus of majority of the studies was on two of the most influential constructs, i.e. perceived usefulness and perceived ease of use. The studies examined the effective part played by the two constructs in ensuring that the students' adoption or acceptance is on-demand [22] , [24] - [26] . UTAUT, which is an extended model of TAM, has also been employed as a model to determine how effective the constructs have been during the pandemic. Different technology acceptance models were used in the studies carried out in India. However, the model was extended by [21] by including SUS, which is vital for examining perceived usability. The same trend was followed in the study by [23] , in which a few external factors were included to extend the TAM model, such as innovativeness, computer self-efficacy, computer anxiety, social norms, perceived enjoyment, content and system quality.

The impact of the pandemic has extended to various parts of the world, which is why the study is varied in its location. Different studies were carried out in Indonesia, China, Vietnam, and Malaysia. In all of these studies, the impact of e-learning platforms during the pandemic was explained with the help of surveys or online questionnaires that were distributed to students of undergraduate educational institutions [24] , [27] - [29] . The studies conducted in Romania and Europe also used surveys or online questionnaires distributed among either students or farmers.

They had a distinct sample because of the distinct objectives of their study. The study in Europe had the objective of describing the farmers' readiness to use new technology during the pandemic, while the one in Romania focused on examining how the online platform affected a sample of students during the pandemic [23] , [30] . The impact of COVID-19 on the educational environme nt of different cities in India was examined by various researchers. It was deduced in these studies that e-learning platforms were effective in maintaining direct and indirect means of communication between the various participants in educational institutions [21] , [22] .

With respect to the methodology, most previous studies used only the single-stage linear data analysis [31] , specifically the SEM technique. It was only possible to identify the linear correlations between the factors in the theoretical model using the single stage of SEM analysis, and this was not sufficient to predict the complex decision-making processes [32] . This limita tio n was addressed by a few scholars by using the ANN method as a second stage of analysis [33] - [35] .

However, this approach comprises only a single hidden layer and is considered as a shallow type of ANN [36] . It has been indicated that the deep ANN architecture should be used rather than the shallow ANN, as it may improve the accuracy of non-linear models by employing over a single hidden layer [37] . Therefore, the present research uses a hybrid SEM-ANN technique that is based on a deep ANN approach which offers deep learning. The previous studies [21] , [22] , [38] - [40] , [23] - [30] have clearly shown that majority of them were carried out in educational institutio ns where the teaching and learning process is mainly carried out through the e-learning platforms. In this way, it ensures that the shift from the traditional classroom to the e-learning environme nt occurs safely and effectively [2] , [3] . In the end, this would enable all the educational institutio ns to accomplish their aims and objectives [26] , [27] .

Within various populations, the COVID-19 vaccination fear and hesitancy is quite high, mainly due to the increase in the coronavirus conspiracy theory. Hence, the vaccination is rejected, and there is a constant increase in the vaccination hesitancy percentage [41] . Generally, the vaccinatio n acceptance is influenced by the risk theory. There are risk regulation and cultural assessments associated with the risk theory. The risk response is quite deep considering the emotions present due to cultural developments. Risk information is evaluated in a way that the expected utility is maximised [42] - [45] . The perceived vaccination fear differs according to different genders. For women, generally, the vaccine is associated with negative experiences regarding the healthcare institution; however, the men's attitude is associated with the immune systems; they believe that their immune systems would be weakened [46] , [47] . Regarding the COVID-19 vaccine, it is observed that vaccine rejection is significantly influenced by the health literacy. Research indicates that the vaccine may be rejected by the students, specifically female students, considering their health literacy. Health-protective behaviour is adopted, since their fear towards the COVID-19

vaccine is quite high. The perceived vaccination fear is what leads to the spread of the COVID-19

infection [48] - [50] . Thus, considering the earlier assessments, the vaccination fear hypothesis was developed.

H1: FV has a positive and significant impact on POA.

The technology acceptance model (TAM) theory was formulated by Fred Davis, contributing to the idea of technology adoption, acceptance and post-acceptance. Perceived ease of use and perceived usefulness are the constructs of this model, and these are taken to be conceptual factors that contribute to the post-acceptance of the technology. The perceived ease of use variable is linked to the effectiveness of the easiness factor on the users' performance, while perceived usefulness is concerned with the idea of "effort-free" that improves the users' performance [51] .

This led to the following hypotheses:

H2: PU has a positive and significant impact on POA.

H3: PE has a positive and significant impact on POA.

The idea of daily routine in this context refers to the degree to which technology is a part of the routine tasks and the inclusion of technology into the standard work routine of users. The daily routine use of technology is described as the use of technology such that it enters into the daily pattern and is considered as a standard element in the life of users [13] , [52] , [53] . The daily routine is a significant factor that affects the post-acceptance model. The effectiveness and the use of outcomes have an impact on the daily routine. This suggests that technology users will consider it as a part of their daily routine if it improves their extrinsic motivation. Additionally, it can support the integration of technology with work processes [13] . However, the impact of daily routine varies for different users, because they may have distinct work situations and views about the integratio n of technology in their routine work [54] . The following hypothesis was hence developed:

H4: PR has a positive and significant impact on POA.

Albert Bandura was the first person to present the idea of self-efficiency, suggesting that the concept was part of the social cognitive theory and was a vital prerequisite for effective learning behavior. Self-efficiency examines the users' perception regarding their ability to perform various tasks and complete these tasks appropriately [55] , [56] . There is a close relationship between elearning systems and the users' self-efficiency in the classroom. The teachers' capability to efficiently and effectively use technology determines the actual teaching practices in classrooms.

Therefore, if teachers have high efficiency, they will be able to perform the desired tasks accurately. This goal can be attained by involving students in different activities, and this will serve as an encouragement for teachers to use technologies more consistently and to slowly enhance their proficiency [57] - [60] . It has been observed during the COVID-19 pandemic that self-efficiency in the educational setting has been influenced by the pandemic situation. It was suggested by various researchers that COVID-19 has had an impact on self-efficiency. According to [61] , during the pandemic crisis, work commitment among teachers has been influenced by selfefficiency. It was also stated by [62] that self-efficiency potentially impacts technology adoption during COVID-19. The authors asserted that the adoption can be curtailed or even avoided when there is awareness regarding the drawbacks of the pandemic. Therefore, self-efficiency may play a protective role during the pandemic, because it may give rise to a more flexible environment that supports technology adoption. It seems that the vaccination is a solution to prevent the pandemic spread; however, several people refuse to take it because of various reasons. One of the main reasons is the lack of trust in the health system [63] - [65] . Considering the fact that the fear of vaccination affects the mental and physical health of the users, the following was hypothesised:

H5: SE has a positive and significant impact on POA.

Csikszentmihalyi presented the flow theory as a way of understanding the motivation of users.

There is a close relationship between motivation and the psychological state in which the cognitive feeling of motivation and efficiency manages the users [66] , [67] . The flow theory signifies the state in which users are excessively involved in a given activity. Users gain a highly enjoyable experience when using the technology; therefore, they will do it in any situation. There is a relationship between flow theory and intrinsic motivation, and particularly self-motivation. It is believed that self-motivation is one of the best ways in which users can be encouraged to do various activities while achieving inner satisfaction. Current research suggests that positive attitudes and high effectiveness is attained from flow. Further, it supports high education objectives achievements since students become motivated to attain specific objectives [68] - [70] . Since the COVID-19 pandemic occurred, various researchers have presented similar results. The motivatio n of the students during the pandemic was found to affect and be indicative of their satisfactio n, success and learning outcomes [71] , [72] . User performance is influenced by the fear of vaccination that is expected to reduce the pandemic's bad effects. Hence, the following was hypothesised:

H6: EJ has a positive and significant impact on POA.

According to the critical mass theory, a significant contribution is made by a population group when specific actions are being adopted. Hence, the behaviour is considered essential by the other individual and the same behaviour is imitated. The critical mass influence for technology adoption is quite vital. If a group of friends has decided to make use of a technology, it is expected that the next group would also adopt it [73] - [75] . Hence, the following hypothesis was formed:

H7: PC has a positive and significant impact on POA. 

The students from UAE universities were given online surveys to fill. The data was gathered during the winter semester 2020/2021, from 25 January 2021 to 30 May 2021. 700 questionnaires were randomly distributed by the research team, out of which 659 questionnaires were filled by the respondents. Hence, the response rate stood at 94%. Since there were a few missing values, 44 of the filled questionnaires were rejected. Therefore, 659 questionnaires were considered to be useful and effective by the team. The sample size of 659 valid responses is quite appropriate as stated by [76] . The projected sampling size for the 1500 population as part of 306 respondents. When compared to the insignificant requests, the sample size of 659 is quite high. Hence, as the sample size, the structural equation modeling would be accepted [77] , and this was needed for hypotheses' confirmation. Additionally, the current theories have been used to establish the hypotheses; yet, they have been included as part of the M-learning context. The measurement model was evaluated using SEM (SmartPLS Version 3.2.7). Advanced treatment was carried out by applying the final path model.

The respondents' personal/demographic data is presented in Table 1 . Among the respondents, 48%

were males and 52% females, with age group ranging between 18 and 29 years. Nearly 34% of the respondents were older than 29 years. Most respondents had a university degree and a sound educational background. 62% held a bachelor's degree, 23% a master's degree, 12% a doctoral degree and the rest held diplomas. The voluntary respondents of the research [78] recommended that a "purposive sampling approach" be used. The study sample was developed using respondents of various ages and from different colleges, programmes or levels. The IBM SPSS Statistics ver.

23 was used to measure the respondents' demographic information. 

The survey is the instrument which was developed to test the hypotheses of the research. There are 23 items part of the survey which assessed eight constructs present within the questionna ire. Table 2 indicates the sources of all eight constructs. Earlier studies' research questions, after being subjected to some modifications, have been included to ensure the research is applicable. 

A pilot study was carried out to check the reliability of the questionnaire items. Accordingly, 70 students were randomly selected from the mentioned population. Considering the total sample size to be 10% for the current analysis, the student sample size of 700 was decided, and the research standards were considered appropriately. To assess the pilot study's findings, the interna l reliability was tested using Cronbach's alpha (CA) test along with the IBM SPSS Statistics ver.

23. Hence, for the measurement items, acceptable conclusions are presented. It is acceptable to have a 0.70 reliability coefficient if the social science research studies mentioned pattern is emphasised upon [79] . Table 3 presents the seven measurement scales Cronbach alpha values. 

The questionnaire survey was circulated by the researcher. Online surveys were received by the students within United Arab Emirates (UAE) universities (N = 700). Two kinds of popular UAE universities were approached within this research.

The questionnaire survey given to the students [78] had three sections.

The first section emphasises upon the respondent's personal data.

The second section included the two items which indicate the general questions regarding e-learning systems.

The third section included 21 items related to "the fear of vaccination, perceived ease of use, perceived usefulness, perceived routine use, perceived enjoyment, perceived critical mass and self-efficiency".

A five-point Likert Scale was used to measure the 23 items with the following indicators: strongly disagree (1), disagree (2), neutral (3), agree (4) and strongly agree (5).

The earlier empirical studies used the SEM to carry out a single-stage analysis. For the current research, a hybrid SEM-ANN approach using deep learning for validation of the hypothesised associations between the research model factors was applied. There are two phases within this research. At first, the recommended research model would use the partial least squares-structura l equation modeling (PLS-SEM) by applying SmartPLS [80] . The objective of using PLS-SEM within the research is the theoretical model exploratory nature and the non-availability of earlier literature. General guidelines were followed by the study for implementing the PLS-SEM within the information systems research [81] . It has already been mentioned [82] that the research model can be analysed using a two-step approach (i.e., measurement model and structural model).

Further, IPMA was applied within this research as it is an advanced PLS-SEM technique that helps extract the performance and importance of each construct within the research model. Second, the PLS-SEM analysis investigation, complement, and authentication would be done through ANN.

It would also help state the independent variables effectiveness upon the dependent variable. The ANN is considered to be an instrument for function approximation, which is relevant in areas where the collaboration between input(s) and output(s) is non-linear as well as complex. ANN states that there are three vital mechanisms: network architecture, learning rule and transfer function [82] . It is then divided into further subcategories such as "radian basis, feed-forward multilayer perceptron (MLP) network and recurrent network" [32] . One commonly used approach is the MLP neural network in which several layers are present, such as input and output. Hidden nodes are used to connect these input and output layers. Neurons or independent variables are present within the input layer, and they are responsible for taking the raw data ahead to the hidden layers as synaptic weights. The activation function choice determines the hidden layer output. A widely used activation function is the sigmoidal function [83] , [84] . The MLP neural network has been used for the recommended research model training and testing.

The measurement model was assessed keeping in mind the construct reliability (which integrates "composite reliability (CR), Dijkstra-Henseler's (PA), and Cronbach's alpha (CA)") along with validity (which integrates convergent and discriminant validity). This recommendation was brought forward by [85] . The construct reliability was determined using Table 4 which presents the CA values of 0.769 till 0.906. As compared to the threshold value of 0.7 [86] , these figures are higher. Table 4 indicates that the CR values are present between 0.786 and 0.927, and these are also greater than the threshold or recommended value of 0.7 [87] . The Dijkstra-Henseler's rho (pA) reliability coefficient should be used by the research for construct reliability evaluation and reporting.

The reliability coefficient ρA should be similar to the CA and CR with values of 0.70 or higher as part of the exploratory research, and for advanced research stages, it should be over 0.80 or 0.90 [86] , [88] , [89] . It has been indicated in Table 4 that for each measurement construct, the reliability coefficient ρA should be over 0.70. As the results suggest, the construct reliability should be confirmed and the constructs must be assumed to be quite error-free towards the end.

The convergent validity measurement testing should be done for the average variance extracted (AVE) and factor loading [85] . The findings in Table 4 indicate that the suggested value of 0.7 remained lower than the factor loading values. Further, the AVE-produced values lay between 0.686 and 0.842, and these are much higher than the threshold value of 0.5. For all constructs, it is possible to appropriately attain the convergent validity keeping the future results in mind.

For the measurement of discriminant, it has been recommended that "the Fornell-Larker criterion and the Heterotrait-Monotrait ratio (HTMT)" be measured [85] . The outcomes in Table 5 indicate that the Fornell-Larker condition confirms the requirements as the AVEs and their square roots are much higher than the correlation with the rest of the constructs [90] . Table 6 illustrates the HTMT ratio results, which indicate the fact that 0.85 is the threshold value which is ahead of each construct's value [91] . Therefore, the HTMT ratio is created. Based on these findings, the discriminant validity is stated. Keeping in mind the analysis outcomes, there were no problems related to the measurement model's assessment in terms of the reliability and validity. Hence, it is possible to assess the structural model by making further use of the collected data. 

The fit measures offered by SmartPLS are "standard root mean square residual (SRMR), exact fit criteria, d_ULS, d_G, chi-square, NFI, and RMS_theta" which shows the PLS-SEM model fit [92] .

Through the SRMR, it is possible to indicate the difference amongst the observed correlations and model implied correlation matrix [93] . The good model fit measures are the values below 0.08 [94] . A good model fit is when the NFI values are greater than 0.90 [95] . The chi-square value ratio is the NFI for the recommended model to the benchmark or null model [96] . The NFI would be highly based on the large parameters, which is why it is not considered to be an appropriate model fit indicator [93] . The discrepancy amongst empirical covariance matrix and the covariance matrix indirect to the composite factor model can be extracted through the two metrics : the squared Eucledian distance d_ULS and the geodesic distance d_G [93] , [97] . The RMS theta can only be applied to the reflective models which assess the outer model residuals' correlation degree [96] .

The PLS-SEM model is considered efficient if the RMS theta is close to zero. The values are a good fit if they are lower than 0.12; but for any other value, it indicates a lack of fit [98] . The correlation amongst the constructs can be evaluated through the saturated model according to [93] ; however, the model structure and total effects are only observed by the estimated model.

According to Table 7 , 0.070 is the RMS_theta value, which indicates that the global PLS model validity can be demonstrated due to the large size of the related goodness-of-fit for PLS-SEM model. 

The Smart PLS, with maximum likelihood estimation, was integrated with the structural equation model to extract the several structural model theoretical constructs interdependence [99] , [100] .

Hence, it was possible to assess the proposed model. It is observed in Table 8 as well as Figure 2 that a moderate predictive power is present for the model, since it attains a 45% variance percentage for post-acceptance of e-learning technology.

The beta (β) values, t-values, and p-values for each hypothesis, considering the PLS-SEM technique generated results, are stated in Table 9 . It can be stated that all hypotheses have been supported by the researchers. Considering the data analysis, the empirical data supports the H1, H2, H3, H4, H5, H6, and H7 hypotheses.

The 

The IBM SPSS Statistics ver. 23 was used to conduct the ANN review. Only the notable predictors were produced from the PLS-SEM findings, which were then used in the ANN evaluation. to take effect for each of the output neuron nodes, a two-hidden layer deep ANN model was used [101] . The sigmoid function was used as the activation function for both hidden and output neurons in this study. To improve the efficiency of the presented research model [102] , the spectrum for both input and output neurons was specified between [0, 1]. A tenfold cross-validation approach with a ratio of 80:20 for both training and testing data was used to prevent overfitting in ANN models [83] . The root mean square of error (RMSE) was proposed as a measure of the neural network model's accuracy. The RMSE parameters of the ANN model for both training and testing data were 0.1394 and 0.1405, respectively, as seen in Table 4 . Since the RMSE estimates and standard deviation for both training and testing data were of minuscule variances(0.0046 and 0.0098, respectively), it can be assumed that the presented research model achieves high accuracy with the use of ANN.

The average of every predictor was compared to the maximum mean value expressed as a percentage to determine the normalised importance. Every one of the predictors involved in ANN modelling is included in Table 5 with their mean importance and normalised importance. The sensitivity analysis results show that PE is by far the most important predictor of post-acceptance of e-learning technology, preceded by FV, PR, EN, SF, PC, and PU, as shown in Table 5 . It was proposed that the goodness of fit, which is identical to R 2 in PLS-SEM analysis [103] , be determined to additionally authenticate and validate the ANN application's accuracy and performance. As an outcome, the predictive power of ANN evaluation (R 2 = 79%) was far greater than that of PLS-SEM (R 2 = 45%), according to the findings. These results show that the ANN model elucidates endogenous constructs more thoroughly than the PLS-SEM approach. Further, the disparity in variances can be due to the deep learning ANN approach's dominance in deciding non-linear relationships between the constructs. 

The IPMA was used as an efficient approach in PLS-SEM with post-acceptance of e-learning technology as the target variable in this study. IPMA, according to Ringle and Sarstedt [104] , aids in the comprehension of PLS-SEM research evidence. IPMA includes the average value of the latent constructs and their indicators (e.g., performance measure) as an option to just measure the path coefficients (e.g., importance measure) [104] . According to IPMA, the overall impact represents the importance of precedent factors in defining the target factor (post-acceptance of elearning technology), while the average of latent constructs' values reflects their performance. The IPMA findings are illustrated in Figure 4 . The importance and performance of the seven variables 

The current analysis looked into the key factors that influence the acceptance of an e-learning platform using the TAM model in conjunction with external factors such as "fear of vaccinatio n, perceived ease of use, perceived usefulness, perceived routine use, perceived enjoyment, perceived critical mass and, self-efficiency". The following are some of the most important findings. First, it was discovered that all of the variables have an impact and are linked to the dependent variable.

This finding backs up the TAM in the study, which shows that the two constructs of "perceived ease of use and perceived usefulness" have a favourable impact on technology acceptance [23] , [105] - [107] . According to a study by [108] , perceived ease of use and perceived usefulness have a significant impact on Zoon's acceptance as an e-learning platform, and they have a successful relationship with self-efficacy. The results of this study show that both PEOU and PU have an important and optimistic effect on the e-learning platform's post-acceptance.

Second, the assumptions concerning "perceived routine use and perceived enjoyment" were ensured. These results are consistent with those observed in earlier research. According to [13] , the perceived routine use has a beneficial effect on technology acceptance and has a direct association with motivation; similarly, prior research has found that perceived enjoyment has an important impact on technology acceptance [23] , [107] , [109] , [110] . According to an analysis by [111] , perceived enjoyment has a favorable impact on technology acceptance, and this factor is closely linked to fast internet and very well systems.

Consequently, the statistical analysis backed up critical mass and self-efficiency. The outcome demonstrates that in moments of distress, people try to be compassionate. They assist one another in avoiding technical difficulties. As a result, these two parameters can act as a buffer througho ut a crisis [61] , [62] , [73] , [74] . According to an analysis by [23] , self-efficacy and TAM (PEOU and PU) are the most important factors that influence acceptance of technology.

Eventually, according to the results, it was found that fear of vaccination is a major variable that influences the acceptance of an e-learning platform. The results show that fear of vaccination has a beneficial effect on the key model constructs. Vaccine hesitancy has been identified as a critical problem, as it can build numerous obstacles in a post-crisis environment [7] - [9] . Students ought to have vaccination confidence as opposed to vaccination hesitancy; as a result, they would be able to build trust in the healthcare sector, allowing them to tolerate vaccinations and overcome their fears.

The fear of vaccination is a significant barrier to students' acceptance of technology. As a result, practical implications can aid in the direction of learning techniques and methods. Fear of vaccination, according to experts, may harm users' views. Therefore, education practitioners and teachers should pay close consideration to students' perceptions and preferences while implementing e-learning models [112] , [113] . Accordingly, this study has several practical implications. To begin, teachers and healthcare workers should explore various methods for boosting the confidence of technology consumers to accept vaccination. Second, to explore this problem further, investigators may obtain practical confirmation of the impact that vaccinatio n fear can have on the learning process. Further, the current research provides a scientific viewpoint.

Fear of vaccination should be taken into account by academics in educational institutions in the online learning environment, particularly in the teaching and learning surroundings. Accordingly, teachers should rethink the importance of evaluating their teaching strategies to address the current obstacles [114] .

This study analysed the impact of vaccination hesitancy in the educational environment to encourage healthcare executives and governments to pay attention to vaccination fear and take appropriate measures to alleviate this fear. It provides an in-depth look at existing strategies for reducing fear and boosting vaccination morale among educators, teachers, and students, which will have an impact on society overall. For example, healthcare managers should accept these results to minimise the risk of a crisis that might end in student vaccination rejection [115] . This research has the potential to improve teaching and learning efficiency in educational institutions as well as increase vaccination acceptance, allowing educational institutions to execute their objectives and perspective more efficiently.

In terms of methodology, contrasting prior empirical studies that mainly depended on SEM analysis, this study used a hybrid SEM-ANN model focused on deep learning to refer to the existing research in general and the e-learning domain in particular. The ANN model has a much higher predictive power than the PLS-SEM model. It is concluded that the higher predictive power extracted from ANN analysis results from the deep ANN architecture's power to assess non-linear associations between the factors in the theoretical model.

There are certain drawbacks to the current research. The key drawback is that only four universities in the UAE were engaged, which prevents a thorough examination of the factors influencing elearning platforms following the propagation of COVID-19. With the involvement of a larger number of universities, the research may have been more useful. By thoroughly analysing the factors that influence e-learning systems, additional analysis would enable an accurate understanding of e-learning systems. Another limitation is the small number of respondents who participated in the study (659 students). According to [78] , the survey questionnaire was developed for data collection. The study could have been improved if a superior instrument and sampling method had been used. Further, the participation of various universities from the Arab Gulf region, including those from the KSA, Kuwait, and Bahrain, would have enhanced study results. In the future, it will be imperative to engage more students in participating in research. Further, interviews and focus groups would provide more reliable results. Finally, we hope that the participating Arab universities would adopt an e-learning system.

The digital/technological connection with Covid-19: An unprecedented challenge in university teaching

COVID-19 and digital disruption in UK universities: Afflictions and affordances of emergency online migration

3: Connection, digita l education, and student-centric teaching practice before COVID-19

EFL students' challenges in English-med ium business programmes: Perspectives from students and content teachers

Starting from the Same Spot: Exploring Mindfulness Meditation and Student Transitions in the Introductory Health Promotion Classroom

College students' use and acceptance of emergency online learning due to Covid-19

Vaccine Hesitancy on Social Media: Sentiment Analysis from

Prevalence and predictors of vaccine hesitancy among expectant mothers in Enugu metropolis, South-east Nigeria

Vaccine hesitancy: clarifying a theoretical framework for an ambiguous notion

Vaccine confidence in the time of COVID-19

Restoring confidence in vaccines in the COVID-19 era

Public fear of vaccination: separating fact from fiction

The nature and determinants of IT acceptance, routinizatio n, and infusion

The diffusion of interactive communication innovations and the critical mass: the adoption of telecommunications services by German banks

Effects of Organizational Climates on the Self-Efficacy of Practitioners in Continuing Higher Education in Korea

Self-efficacy mechanism in human agency

A technology acceptance model for empirically testing new end-user information systems: Theory and results

Flow theory and research

Student engageme nt in high school classrooms from the perspective of flow theory

Perceived critical mass and collective intention in social media-supported small group communication

Perceived usability evaluation of Microsoft Teams as an online learning platform during COVID-19 using system usability scale and technolo gy acceptance model in India

Impact of Information Sharing on Adoption and User Satisfaction among the Wearable Device Users

Commonly Used External TAM Variables in e-Learning, Agriculture and Virtual Reality Applications

Using an extended Technology Acceptance Model to understand students' use of e-learning during Covid-19: Indonesian sport science education context

Highlighting E-learning Adoption Challenges using data Analysis Techniques: University of Kufa as a Case Study

Perception And Intention To Use Elearning From Students' Point of View-An Evidence From Malaysia Local University

APPLICATION OF TECHNOLOGY ACCEPTANCE MODEL (TAM) TOWARDS ONLINE LEARNING DURING COVID-19 PANDEMIC: ACCOUNTING STUDENTS PERSPECTIVE

The e-Learning Quality Model to Examine Students' Behavioral Intention to Use Online Learning Platform in a Higher Education Institution

Students' adoption of e-learning in emergency situation: the case of a Vietname se university during COVID-19

Online Education During the COVID19 Pandemic: Perceptions and Expectations of Romanian Students

A PLS-SEM neural network approach for understanding cryptocurrency adoption

Understanding and predicting the motivators of mobile music acceptance-a multi-stage MRA-artificial neural network approach

Predicting the determinants of the NFC-enabled mobile credit card acceptance: A neural networks approach

Factors affecting retailer's adopti on of mobile payment systems : A SEM-neural network modeling approach

Evaluating the Impact of Knowledge Management Factors on M-Learning Adoption: A Deep Learning-Based Hybrid SEM-ANN Approach

MtNet: a multi-task neural network for dynamic malware classification

A mothed of improving identification accuracy via deep learning algorithm under condition of deficient labeled data

An investigation of the influencing factors of user learning motivation for using the online learning platform of the foreign language school at Baise Universityn

Examining the moderating effect of perceived benefits of maintaining social distance on e-learning quality during COVID-19 pandemic

Attitude towards Adoption of Online Classes during COVID 19: Integrating UTAUT Model with Perceived Cost

COVID-19 vaccine hesitancy is associated with beliefs on the origin of the novel coronavirus in the UK and Turkey

Risk by choice: Regulating health and safety in the workplace

Private choices and public health: The AIDS epidemic in an economic perspective

Social benefit versus technological risk

How can we know the risks we face? Why risk selection is a social process 1

The Role of Gender and Other Socioeconomic Factors in the Adoption of the Contagious Bovine Pleuropneumo nia Vaccine: A Literature Review

Differences in HIV vaccine acceptability between genders

COVID-19: health literacy is an underestimated problem

Interdisciplinary perspectives on health literacy research around the world: more important than ever in a time of COVID-19

Impact of Health Literacy on Fear of Covid-19, Protective Behavior, and Conspiracy Beliefs: University Students' Perspective

A technology acceptance model for empirically testing new end-user information systems: Theory and results

Defining information technology acceptance: A human-centered, management-oriented perspective

Technology use on the front line : how information technology enhances individual performance

Measuring user participation, user involvement, and user attitude

Self-efficacy: toward a unifying theory of behavioral change

Computer self-efficacy: Development of a measure and initial test

Contextual effects on the self-perceived efficacy of high school teachers

Conditions for classroom technolo gy innovations

Tracing teachers' use of technology in a laptop computer school: The interplay of teacher beliefs, social dynamics, and institutional culture

Adoption and Use of Learning Management Systems in Education: The Role of Playfulness and Self-Management. Sustainability 2021, 13, 1127

Crisis Self-Efficacy and Work Commitment of Education Workers among Public Schools during COVID-19 Pandemic

Design and psychometric analysis of the COVID-19 prevention, recognition and home-management self-efficacy scale

Implicit racial/ethnic bias among health care professionals and its influence on health care outcomes: a systematic review

Pandemics and health equity: lessons learned from the H1N1 response in Los Angeles County

The public's role in COVID-19 vaccination: Human-cente red recommendations to enhance pandemic vaccine awareness, access, and acceptance in the United States

Csikszentmihalyi, M. Beyond boredom and anxiety

Flow: The psychology of optimal experience

Flow in computer-mediated communication: Electronic mail and voice mail evaluation and impacts

Targeting motivation-adapting flow theory to instructio na l design

A Serial-Mediation Model to Link Entrepreneurship Education and Green Entrepreneurial Behavior: Application of Resource-Based View and Flow Theory

Impact of podcasting on student motivatio n in the online learning environment

Effects of augmented reality-based multidimensional concept maps on students' learning achievement, motivation and acceptance

The influence of perceived convenience and curiosity on continuance intention in mobile English learning for high school students using PDAs

Post-adoption switching behavior for online service substitutes: A perspective of the push-pull-mooring framework

Perceived critical mass effect on groupware acceptance

Determining sample size for research activities

Sample size estimation using Krejcie and Morgan and Cohen statistical power analysis: A comparison

Students' Attitudes Towards the Use of Mobile Technologies in e-Evaluation

Psychometric theory

SmartPLS 3. Bönningstedt: SmartPLS

PLS-SEM in Information Systems Research: A Comprehensive Methodological Reference

Artificial neural systems

Examining the role of trust and quality dimensions in the actual usage of mobile banking services: An empirical investigation

An integrated SEM-Neural Network approach for predicting determinants of adoption of wearable healthcare devices

An updated and expanded assessment of PLS-SEM in information systems research

Psychometric theory

Principles and practice of structural equation modeling

PLS-SEM: Indeed a silver bullet

The use of partial least squares path modeling in international marketing

Evaluating Structural Equation Models With Unobservable Variables and Measurement Error

A new criterion for assessing discrimina nt validity in variance-based structural equation modeling

Model Fit

A primer on partial least squares structural equation modeling (PLS-SEM)

Fit indices in covariance structure modeling: Sensitivity to underparameterized model misspecification

Significance tests and goodness of fit in the analysis of covariance structures

Latent variable path modeling with partial least squares

Consistent and asymptotically normal PLS estimators for linear structural equations

Common beliefs and reality about PLS: Comments on Rönkkö and Evermann

An empirical examination of continuo us intention to use m-learning: An integrated model

Explor ing Students' Acceptance of E-Learning Through the Development of a Comprehens ive Technology Acceptance Model

Wearable payment: A deep learning-based dual-stage SEM-ANN analysis

Predicting the determinants of mobile payment acceptance: A hybrid SEM-neural network approach

A hybrid SEM-neural network analysis of social media addiction

Gain more insight from your PLS-SEM results

An Integrated Model of Continuous Intention to Use of Google Classroom

Teachers' Perceptions and Readiness toward the Implementation of Virtual Learning Environment

Factors that affect e-learning platforms after the spread of COVID-19: post acceptance study

Measuring Students' Use of Zoom Application in Language Course Based on the Technology Acceptance Model (TAM)

An analysis of learners' intentions toward virtua l reality online learning systems: a case study in Taiwan

Determinants predicting undergraduates' intention to adopt e-learning for studying english in chinese higher education context: A structural equation modelling approach

Factors affecting students' intentions to undertake online learning: an empirical study in Vietnam

Social patterning and stability of intention to accept a COVID-19 vaccine in Scotland: Will those most at risk accept a vaccine?

COVID-19 and Higher Education: First-Year Students' Expectations toward Distance Learning

K-12 teachers' remote teaching selfefficacy during the pandemic

Anxiety, sleep disorders and self-efficacy among nurses during COVID-19 pandemic: A cross-sectional study