A PLS-SEM Neural Network Approach for Understanding Cryptocurrency Adoption PDF

Received November 26, 2019, accepted December 8, 2019, date of publication December 16, 2019, date of current version January 22, 2020. Digital Object Identifier 10.1109/ACCESS.2019.2960083 A PLS-SEM Neural Network Approach for Understanding Cryptocurrency Adoption OSAMA SOHAIB 1 , WALAYAT HUSSAIN...

Received November 26, 2019, accepted December 8, 2019, date of publication December 16, 2019, date of current version January 22, 2020. Digital Object Identifier 10.1109/ACCESS.2019.2960083 A PLS-SEM Neural Network Approach for Understanding Cryptocurrency Adoption OSAMA SOHAIB 1 , WALAYAT HUSSAIN 1, MUHAMMAD ASIF 2, MUHAMMAD AHMAD 3, AND MANUEL MAZZARA 4 1 School of Information, Systems, and Modelling, Faculty of Engineering and IT, University of Technology Sydney, Ultimo, NSW 2007, Australia 2 Department of Computer Science, National Textile University, Faisalabad 37610, Pakistan 3 Department of Computer Engineering, Khwaja Freed University of Engineering and Information Technology, Punjab 64200, Pakistan 4 Institute of Software Development and Engineering, Innopolis University, 420500 Innopolis, Russia Corresponding author: Muhammad Asif ([email protected]) ABSTRACT The majority of previous research on new technology acceptance has been conducted with single-step Structural Equation Modeling (SEM) based methods. The primary purpose of the study is to enhance the new technology acceptance based research with the Artificial Neural Network (ANN) method to enable more precise and in-depth research results as compared to the single-step SEM method. This study measures the relation between technology readiness dimension (optimism, innovativeness, discomfort, insecurity) and the technology acceptance (perceived ease of use and perceived usefulness) – and the intention to use cryptocurrency, such as bitcoin. The contribution of this study include the use of a multi- analytical approach by combining Partial Least Squares- Structural Equation Modeling (PLS-SEM) and Artificial Neural Network (ANN) analysis. First, PLS-SEM was applied to assess which factor has significant influence toward intention to use cryptocurrency. Second, an ANN was employed to rank the relative influence of the significant predictor variables attained from the PLS-SEM. The findings of the two-step PLS-SEM and ANN approach confirm that the use of ANN further verifies the results obtained by the PLS-SEM analysis. Also, ANN is capable of modelling complex linear and non-linear relationships with high predictive accuracy compared to SEM methods. Also, an Importance-Performance Map Analysis (IPMA) of the PLS-SEM results provides a more specific understanding of each factor’s importance-performance. INDEX TERMS Bitcoin, cryptocurrency, neural network, PLS-SEM, technology readiness. I. INTRODUCTION cryptocurrencies. However, there is scant scholarly research Cryptocurrencies such as bitcoin and Blockchain technology on technology readiness associated with digital money. Bit- have gained significant attention in recent decade. It had coin reached a maximum price in December 2017 when it was been anticipated that cryptocurrencies would have a disrup- valued at over USD$15,000. However, the value of bit- tive effect on financial systems. Even though there has coin decreased since 2017, which damaged consumer enthu- been an increase in not only the economic implications of siasm. The questions that emerge are: what led to this decline cryptocurrencies and degree of interest in this, academic stud- in the interest of individual consumers in cryptocurrency? ies on blockchain-based cryptocurrencies have only recently And why didn’t the consumer adoption of cryptocurrency con- surfaced –. The important issues concerning cryp- tinue? This paper addresses the technology readiness factors tocurrency economics and investment decision-making are which influence the acceptance of bitcoin. related to the pricing mechanism. However, the use of cryp- Three factors that obstruct the use of bitcoin: i) inadequate tocurrencies is rather limited. infrastructure; ii) possible issues within the bitcoin network, It was generally believed that bitcoin might eventually and iii) being apprehensive of the unknown. Technology become a mainstream currency when there was an increase in can support complicated financial dealings as well as mone- consumer demand. Much has been written in the media about tary transfers across borders. Previous research suggests that technology acceptance is impacted by an individual’s per- The associate editor coordinating the review of this manuscript and sonality and demographics ,. The technology readi- approving it for publication was Valentina E. Balas. ness index (TRI) measures a consumer’s readiness to accept This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see http://creativecommons.org/licenses/by/4.0/ 13138 VOLUME 8, 2020 O. Sohaib et al.: PLS-SEM Neural Network Approach for Understanding Cryptocurrency Adoption new technology. TR is measured by evaluating the attitudes an unknown group called Satoshi Nakamoto and was referred and perspectives of consumers regarding technology. to as ‘‘the people’s currency’’. Bitcoin relies on blockchain Such as, optimism and innovativeness have a positive impact, technology. Bitcoin is an electronic, peer-to-peer cash system which are the motivators to technology adoption. Insecurity that has been developed as an alternative payment mecha- and discomfort suppress the use of new technology. nism, autonomous of central banks, governments and other There exists many theoretical models that helps to under- aspects of the conventional monetary system. A public- stand individual behaviors towards using new technol- private key system is used for data encryption. As with any ogy. However, TAM is the most widely accepted peer-to-peer mechanism, its value shows network externality. model in explaining individual behavioral intention toward This means that the more individuals utilizing the system, the technology usage ,. Furthermore, the use of the greater the value of the system for each user. Conse- personality traits factors such as TR is an important extension quently, the worth of bitcoin is determined by its transaction to TAM toward new technology acceptance , –. ability which is a result of public acceptance. The tech- Therefore, this research investigates the integration of nology acceptance model , states that an individual’s TR personality trait dimensions (discomfort, innovativeness, intention to use is determined by how they view technology, optimism, and insecurity) and TAM (perceived ease of use and this is determined by external conditions, for example, perceived usefulness) to examine their impact on intention to social norms and information availability. use cryptocurrency such as bitcoin. From the years, the researchers are focused on the tech- Therefore, this study adopted the TRAM (TRI and nological and economic features of bitcoin, such as the ver- TAM) model , , to predict the use intention ification of transactions as discussed in , , and the of cryptocurrency such as Bitcoin. However, majority of consumers also have concerns about privacy issues in addi- the previous research findings on TAM and TR has been tion to above-discussed features of bitcoin. In the light of reported using structural equation modelling (SEM) method above discussions, the researchers are now focusing their (e.g. linear relations between constructs). So, the primary attention on user adoption. For example, Tsanidis et al. aim of this paper is therefore to enrich the new technol- examined Greek consumers’ awareness of bitcoin, its use ogy acceptance based research with the help of two-step and their degree of trust, finding that prospective users were approach, which is Partial Least Squares- Structural Equation not familiar with the information on bitcoin, for example, Modeling (PLS-SEM) and Artificial Neural Network (ANN) its usefulness, ease of use and other potential advantages. analysis. ANN is capable of modelling complex linear and Bohr and Bahir conducted an online survey in 2013 and non-linear relationships with high predictive accuracy com- found that the average user is aged 32.1 years. They also pared to SEM methods –. In addition, according found that that anonymity (approximately 8% of the sam- to Henseler et al. and Hair et al. , PLS-SEM per- ple), inadequate trust in the banking system (approximately forms better than covariance-based (CB) SEM in finding the 10% of the sample) and freedom (approximately 16% of true model. Therefore, the study employs a multi-analytical the sample). Silinskyte studied individual bitcoin usage approach by combining Partial Least Squares- Structural behaviour based on the Unified Theory of Acceptance and Equation Modeling (PLS-SEM) and ANN analysis. the Use of Technology (UTAUT) model. The survey The paper is organized as follows. Section 2 presents found that the factors that had an impact on bitcoin usage the theoretical background of cryptocurrency and technol- included effort and performance expectancy, and behavioural ogy readiness. In section 3, we discuss hypotheses develop- intention. ment. Following this, in section 4, we describe our research methodology. Section 5 presents the results of this study. B. TAM Section 6 discusses the findings and implications of the study. Technology Acceptance Model (TAM) is the most Finally, this study concludes with the limitations and future widely accepted model to predict technology adoption. TAM directions for research. states that perceived ease of use (PEOU) and perceived use- fulness (PU) have an influential effect on the acceptance and II. RELATED STUDIES AND THEORETICAL BACKGROUND actual use of the technology. A. BLOCKCHAIN AND CRYPTOCURRENCIES Previous researchers have acknowledged the robustness A blockchain is a shared database that enables users to of TAM and extended the framework with external fac- perform transactions of valuable assets within a public and tors significant to technology use. However, the TAM pseudonymous system without depending on a mediator was initially developed to predict technology acceptance in or central body ,. There are three generations of work environments, researchers such as , , , blockchain technology development: Blockchain 1.0, 2.0 and extended TAM by integrating individual-specific constructs 3.0 named as digital currency, digital economy and digital of TR. society, respectively. Cryptocurrency is the most commonly used operational C. TECHNOLOGY READINESS (TR) blockchain mechanism. The most commonly used type of Addressing cryptocurrency adoption concerns such as user cryptocurrency is bitcoin. Bitcoin was developed in 2008 by trust and privacy issues should lead to mature user adoption. VOLUME 8, 2020 13139  O. Sohaib et al.: PLS-SEM Neural Network Approach for Understanding Cryptocurrency Adoption Cryptocurrencies have not yet gained mainstream user adop- tion, being considered a fairly recent innovation. According to McDougall , cryptocurrency is still in its initial phase of adoption. As cryptocurrencies are highly innovative and technology-intensive, technology readiness theory is appro- priate to investigate cryptocurrency user adoption. Technology readiness (TR) refers to an individual’s overall state of mind in terms of technology belief and attitude. The Technology Readiness Index (TRI) scale formulated by Para- FIGURE 1. Research model. suraman determines the extent to which an individual is ready to adopt the technology. Previous studies on the individ- ual use of new technologies suggest that consumers’ beliefs, is very easy, and unlike to focus on any negative conse- perceptions, feelings, and motivation can simultaneously be quences. According to Walczuch et al. optimists favorable (drivers) as well as unfavorable (inhibitors) in terms fearless about negative outcomes. Therefore, optimism has of high-tech products and services. Customers with extremely a positive effect on both the ease of use (PEOU) and per- positive attitudes regarding technology show greater accep- ceived usefulness (PU) of a given technology , ,. tance of technological products and services. In contrast, con- Therefore, optimists have more positive attitudes towards sumers with extremely negative attitudes towards technology technology use, so it is assumed that an optimist perceives are hesitant to adopt technology-related services or products. the adoption of cryptocurrency as easy to use and useful. Parasuraman and Colby segregate technology adop- H1a: Optimism is positivity related to perceived ease of use tion into four different extents: innovativeness, optimism, toward use intention of cryptocurrency such as bitcoin. discomfort and insecurity. Two aspects, innovativeness and H1b: Optimism is positivity related to perceived usefulness optimism have a positive impact on technology readiness, toward use intention of cryptocurrency such as bitcoin. making individuals more inclined to use new technology. Innovativeness reflects the tendency of the user to become Other two aspects discomfort and insecurity impede in the a pioneer in the technological domain. Innovative way of technology readiness, deferring or restricting new adopters are risk-takers and enjoy trying new things. technology endorsement. Lam et al. assert that these Karahanna et al. revealed that innovative individuals four aspects of TRI have a significant impact on technology beliefs about technology adoption are less arduous and more acceptance, hence each should be considered as a predictor innovative individuals are the prime embracers of technology. of adopting technology-based services or products. The study Previous researches have also reported a positive influence of carried out by Lin et al. that concentrated on technology innovativeness on both PU and PEOU , , ,. embracement by linking online services to the customers. H2a: Innovativeness is positively related to perceived ease There is an intuitive correlation between the factors relevant of use toward use intention of cryptocurrency such as bitcoin. for the technology acceptance model and technology readi- H2b: Innovativeness is positivity related to perceived ness ,. Ratchford and Barnhart argue that usefulness toward use intention of cryptocurrency such as TRI influence various cognitive and affective constructs (for bitcoin. example, anxiety, fun, enjoyment, confusion and frustration) Discomfort reflects a feeling of being overwhelmed by related to technology adoption. technology and a perceived lack of control over technol- ogy ,. Individuals who experience a high degree III. HYPOTHESES DEVELOPMENT of discomfort regarding new technologies usually find it Parasuraman and Colby developed TRI 2.0, which is difficult to use technology. Similarly, discomfort leads a robust predictor to measure actual and technology related to difficulties in accepting new technologies , behavioral intentions. The TRI model provides a theoretical and indicates the apprehensions and concerns of users foundation to determine the motivators and inhibitors of new when using technology-related services or products. There- technology acceptance. In this study, the effect of TR (such fore, discomfort harms PU and PEOU of a given technol- as optimism, innovativeness, discomfort, and insecurity) on ogy , , ,. TAM (perceived ease of use perceived usefulness) are exam- H3a: Discomfort is negatively related to perceived ease of ined. The adapted TRAM (TRI and TAM) is used to predict use toward use intention of cryptocurrency such as bitcoin. the use intention of cryptocurrency such as Bitcoin. The TR H3b: Discomfort is negatively related to perceived useful- and TAM integration has been studied by researchers such ness use intention of cryptocurrency such as bitcoin. as , , , ,. Figure 1 illustrates the research Insecurity in terms of technology refers to uncertainty and model. a lack of trust related to security and privacy , ,. Optimism imitates an affirmative perspective of technol- According to Son and Han , insecurity is considered to be ogy that motivates and recommended the technology adop- an inhibitor of technology readiness. It is likely that insecure tion which leads to productivity and flexibility. Those users will be uncertain about new technology and may not with a high level of optimism perceive the use of technology be willing to make an effort to determine whether or not it is 13140 VOLUME 8, 2020 O. Sohaib et al.: PLS-SEM Neural Network Approach for Understanding Cryptocurrency Adoption beneficial to them. Therefore, insecurity has a negative effect Model (SEM) is utilized to test the hypotheses using Smart- on PU and PEOU , , ,. PLS V3.2. An effective overview of variance and H4a: Insecurity is negatively related to perceived ease of covariance-based SEM (CB-SEM) is provided by. use toward use intention of cryptocurrency such as bitcoin. In business information systems research, PLS-SEM H4b: Insecurity is negatively related to perceived useful- is a preferable approach to analyse statistical data due ness toward use intention of cryptocurrency such as bitcoin. to several reasons, i.e. small sample size, does not Finally, Research has proposed that ease of use effects involve normality, able to work without distributional usefulness of a given technology. Furthermore, extensive assumptions with nominal, ordinal and interval-scaled fac- research has suggested the significant effect of ease of use and tors ,. Henseler et al. reviewed the work of perceived usefulness on usage intention , , ,. Rönkkö and Evermann and showed that PLS-SEM per- Therefore, it is proposed forms better than CB-SEM in finding the true model. H5: Perceived ease of use is positively related to perceived Furthermore, Hair et al. emphasized that PLS-SEM usefulness toward use intention of cryptocurrency such as is significantly better than CB-SEM in explaining variance bitcoin. in the dependent factor indicators. Besides, either we are H6: Perceived ease of use is positively related to use inten- working with reflective or formative measurement model, tion of cryptocurrency such as bitcoin. PLS has the advantage to examine data with no bias from H7: Perceived usefulness is positively related to use inten- composite model. However, according to Kock , tion of cryptocurrency such as bitcoin. Variance Inflation Factor (VIFs) are the indicators to test the biases in data. The threshold value is 3.3 for a full collinearity IV. RESEARCH METHODOLOGY test. If the results are less than equal to threshold, then the In this study, a multi analytical methodology is employed by model is unbiased. In our research model, all VIFs are lower integrating Partial Least Squares Structural Equation Mod- than 3.3, indicating no bias in the data. elling (PLS-ESM) with a most significant artificial intel- Moreover, Henseler and Sarstedt showed that model ligence technique named as Artificial Neural Network fit indices such as goodness-of-fit (GoF) and the relative (ANN). The analysis is performed in two phases. First goodness-of-fit index are not appropriate for model validation phase is related to PS-SEM, which is further divided in two in the PLS approach. PLS-SEM is now a well-established steps named as measurement model validation and structural method in information systems research , Therefore, model hypotheses testing. Henseler et al. stated: ‘‘In variance-based SEM (also called component-based SEM) is research settings with predictive scope, weak theory, and appropriate for this study. In our research model, all factors no need for an understanding of underlying relationships, were modelled as reflective indicators. artificial neural networks (ANN) may be useful’’. In the second phase, ANN is applied to examine the com- A. MEASUREMENT MODEL plement and verify the PLS-SEM analysis and measure the The measurement model includes two assessments of validity effectiveness of independent factors on the dependent fac- and reliability, which are measured by investigating inter- tor. The two methods are explained in detail in section 5. nal consistencies, convergent and discriminant validity. Our research method is aligned with previous research such Cronbach’s reliability and internal consistencies with com- as , , ,. posite reliability for each latent factor exceed the recom- A data collection survey was conducted in Australia mended value of 0.70. Figure 2 shows that the loading of in 2019. We contacted graduate and undergraduate students all items for each reflective construct exceeded the value and staff at the University of Technology Sydney. A Lik- of 0.7 and was significant (p-value < 0.05). ert scale based closed-ended questionnaire is prepared to collect the responses. (1) Strongly Disagree, (2) Disagree, (3) Neutral, (4) Agree and (5) Strongly Agree. The reliability and validity of measurement scales are ensured by modify- ing previously utilized instruments to ensure survey validity. Appendix A details all the items used in the study. TRI is adopted and modified from. TAM (perceived ease of use and perceived usefulness) is modified from. V. DATA ANALYSIS AND RESULTS A total of 160 participants, 56% males and 44% females, completed the survey. The incomplete responses were removed out, and 140 are utilized for further analysis. All the participants are well aware of blockchain technol- ogy and cryptocurrencies. Variance based statistical analysis model called Partial Least Square (PLS) Structural Equation FIGURE 2. Structural model results. VOLUME 8, 2020 13141  O. Sohaib et al.: PLS-SEM Neural Network Approach for Understanding Cryptocurrency Adoption TABLE 1. Reliability and validity assessment. TABLE 3. Structural model testing. TABLE 2. Discriminant validity- HTMT. usefulness. Furthermore, discomfort and insecurity nega- Table 1 and 2 show the measurement model assess- tively influence the perceived ease of use and perceived ment results. The AVE of all variable values exceeds the usefulness. In addition, perceived ease of use and perceived recommended value of 0.50. Hair recommended that not usefulness positively affect cryptocurrency use intention. to rely on cross-loadings for discriminant validity but instead To measure the cross-validated redundancy, which assesses rely on the Heterotrait-monotrait (HTMT) criteria developed the predictive relevance: Q2 Stone-Geisser criterion is by Henseler et al.. HTMT achieve better discriminant investigated using blindfolding method. Q2 values validity results as compared to the cross-loading in PLS- (i.e. intention to use cryptocurrency = 0.308) is above the SEM. Table 2 shows all HTMT values are below the threshold value of zero, hence representing a strong predictive recommended value of 0.85. relevance. Furthermore, to demonstrate the predictive rele- vance, the PLSpredict algorithm is used to predict the PLS B. STRUCTURAL MODEL model’s performance for the Manifest Variables (MV) and The path coefficients significance was assessed by applying the Latent Variables (LV) ,. The PLSpredict algo- T-test which was computed using the bootstrapping technique rithms involve cross-validated case-wise and average-case and the significance level was 5%. Bootstrapping is a non- point predictions; Mean Absolute Error (MAE), Root Mean parametric method to test the coefficients i.e. path coeffi- Square Error (RMSE) and Mean Absolute Percentage Error cients, outer factor weights by assessing the standard error for (MAPE). The PLSpredict rests on the k-fold cross-validation estimation. SmartPLS V3.2 is utilized to execute both inner principle, which is also useful for the holdout sample valida- and outer model to specify the t-value for significance. The tion. The analysis uses the ten number of folds (k = 10) threshold values for significance level 10%, 5% and 1% are and ten repetitions (r = 10) to perform the PLSpredict esti- 1.65, 1.96 and 2.58 respectively. mation. The prediction error is the RMSE, averaged over all The structural model is assessed by the path coefficients k folds. PLSpredict offers two naïve benchmarks 1) lin- significance and the R square (R2 ) variance of the depen- ear model (LM) predictions and 2) mean value Q2 to measure dent construct. Figure 2 shows the structural model results. the predictive quality of the PLS path model estimations. The results of the R2 indicate that 52% (PU), 60% (PEOU) Table 4 summarises the PLSpredict performance of the latent and 44% of the variance is the cryptocurrency use intention variable (intention to use cryptocurrency) and its manifest (CUSE). The result of the R2 shows a satisfactory level of variables (three items). explanation. The lower values of PLS-SEM compared to the sim- Figure 2 and Table 3 shows the hypotheses testing. The ple linear model (LM) values indicate higher predictive findings show that optimism and innovativeness significantly power. Q2 values are also greater than zero. This shows and positively influence perceived ease of use and perceived the PLS-based prediction yields more accurate out-of-sample 13142 VOLUME 8, 2020 O. Sohaib et al.: PLS-SEM Neural Network Approach for Understanding Cryptocurrency Adoption TABLE 4. PLS Predict results. FIGURE 3. ANN model-1. predictions (i.e., smaller predictions errors) for all indica- tors. All PLS-SEM methods achieve somewhat better results than multiple linear regression (Cryptocurrency Use RMSE from the above authors, the prediction accuracy of the trained 0.581 for PLS). However, even better prediction could be network is measured using ten-fold cross-validation. To avoid achieved with larger samples. overfitting problem data is divided into two parts, from which 90% for training and 10% for testing , ,. C. ANN ANALYSIS The prediction accuracy of the ANN model was computed As discussed in the methodology section, Artificial Neu- by the root mean square error for both the training (90%) ral Network analysis is used in the second phase of the and testing (10%) data sets (ten runs). The RMSE is analysis. Significant hypothesized predictors are utilized as calculated using equation 1 and 2 , where SSE is the sum inputs to ANN to emphasize the relevant importance of of squared error, and MSE is the mean squared prediction each predictor’s variable. The relationship (linear or nonlin- error. ear) between the predictor and adoption decision variables MSE = [1/n] × SSE (1) can also be examined with ANN ,. Also, ANN √ RMSE = MSE (2) produces more precise predictions compared to the SEM approaches ,. SEM analysis could lead to an over- As shown in Table 5 to 7, the RMSE values for the training simplification of the complexities of the decision-making data set and the testing data set to represent an accurate ANN process , ,. On the other hand, the ANN method model in taking the relationships between predictors and the is not recommended for testing hypotheses involving causal output. According to , lower RMSE values represent relationships ,. However, ANN provides a higher higher predictive accuracy and better data fit. prediction accuracy than SEM. Therefore, the use of the PLS-SEM–ANN method in this study would complement TABLE 5. RMSE values for the ANN model-1. each other. In this research, a Multilayer Perceptron (MLP) back prop- agation feedforward method is adopted. The MLP is the most commonly used and popular ANN method ,. The ANN analysis comprises three layers: the input layer, the hidden layer, and the output layer. In our research model, MLP- ANN is modelled using SPSS v22. The PLS-SEM model is decomposed into three ANN models with one output variable. Model 1 (Output - PEOU) and has four inputs Optimism, Innovativeness, Discomfort and Insecu- rity. Model 2 (Output - PU) has five inputs perceived ease of use, Optimism, Innovativeness, Discomfort and Insecu- rity. Model 3 (Output - Cryptocurrency Use Intention) has two inputs perceived ease of use and perceived usefulness. The ANN model-1 is shown in Figure 3. Following rec- ommendations are given by , , , the hidden Moreover, the relative importance of each input predic- neurons (nodes) are automatically generated and activation tor (all three ANN models) was computed in terms of nor- function (Sigmoid Function) is utilized for both hidden and malized relative importance ranking (expressed as a %) output layers. Furthermore, based on the recommendations using sensitivity analysis as presented in Table 8 to VOLUME 8, 2020 13143  O. Sohaib et al.: PLS-SEM Neural Network Approach for Understanding Cryptocurrency Adoption TABLE 6. RMSE values for the ANN model-2. TABLE 9. Normalized variable relative importance (Output: PU). TABLE 10. Normalized variable relative importance (output: CUSE). TABLE 7. RMSE values for the ANN model-3. FIGURE 4. Normalized importance (Output variable: perceived ease of use). TABLE 8. Normalized variable relative importance (output: PEOU). 10 and Figure 4, 5 and 6. Based on the normalized vari- able importance, optimism is the most significant predictor of intention to use cryptocurrency, followed by innovative- FIGURE 5. Normalized importance (Output variable: perceived ness, while discomfort has a weaker influence followed by usefulness). insecurity. Table 11 to 12 compares the results of ANN (all three models) and the PLS-SEM based on the strength of path in ANN and ranked four in PLS-SEM. The reason is that coefficients (PLS-SEM) and normalized relative impor- ANN measure both linear and non-linear relationship among tance (ANN) ranking. The comparison Table 11 (Output: variable with high predictive accuracy –. PEOU-Perceived Ease of Use) show that optimism and inno- The comparison Table 12 (Output: PU-Perceived Useful- vativeness ranked one and two respectively both in ANN ness) show that perceived ease of use, discomfort and insecu- and PLS-SEM analysis. However, discomfort is ranked at rity are ranked similar both in ANN and PLS-SEM analysis. number four in ANN and number three in PLS-SEM in terms However, optimism is ranked higher than innovativeness in of predictor’s influence. Similarly, insecurity is ranked three the ANN analysis. As discussed above, ANN measure both 13144 VOLUME 8, 2020 O. Sohaib et al.: PLS-SEM Neural Network Approach for Understanding Cryptocurrency Adoption TABLE 13. Comparison between PLS-SEM and ANN analysis (output: CUSE). FIGURE 6. Normalized importance (Output variable: perceived usefulness). models. The two-step PLS-SEM and ANN method provided TABLE 11. Comparison between PLS-SEM and ANN analysis (output: better in-depth results regarding the relative importance of the PEOU). input factors, thus representing useful information regarding the new technology use. The PL-SEM analysis shows that optimism has the strongest positive influence on perceived ease of use. ANN analysis confirms these findings, ranking optimism higher than innovativeness. The PLS-SEM results also show that discomfort has a negative influence perceived ease of use, followed by insecurity, but the ANN model predicts that inse- curity has a higher impact than discomfort, which is by far the weakest predictor. The findings from both the PLS and ANN also showed perceived ease of use has the strongest influence on perceives usefulness. The PLS analysis shows innovative- TABLE 12. Comparison OF ANN and PLS- ANN analysis (output: PU). ness is ranked higher than optimism, but this is not the case in ANN. Optimism is ranked higher than innovativeness in the ANN analysis. Discomfort and insecurity are ranking are matched. Finally, both the PLS and the ANN analysis shows both perceived usefulness has a higher significant positive effect on cryptocurrency use intention than perceived ease of use. This study has shown that technology readiness has a significant relationship with user adoption of cryptocurrency such as bitcoin. This study confirms that technology readiness has a significant relationship with technology acceptance (perceived ease of use and perceived usefulness). This shows that optimists and innovative peopl are more willing to try linear and non-linear relationship among variable with high new things and have a more positive attitude towards new predictive accuracy. technology use such as cryptocurrency adoption and use. Finally, the comparison Table 13 (Output: CUSE- However, the two other dimensions of technology readiness Cryptocurrency Use Intention) show that perceived useful- (discomfort and insecurity) suggesting that greater complex- ness and perceived ease of use are ranked one and two ity in using a technology-related product or service leads to respectively both in ANN and PLS-SEM analysis. uncertainties and difficulties in accepting new technologies. Given that Parasuraman and Colby demonstrated that VI. DISCUSSIONS – IMPLICATIONS - CONCLUSION TRI is a robust predictor of technology-related behavioural This papers extends the new technology acceptance-based intentions as well as actual behaviours, the findings obtained research with ANN approach, which is traditionally based on in this study are consistent with the hypotheses. According to SEM technique. The strength of each predictor input to the these authors, the dimensions optimism and innovativeness output (perceived ease of use, perceived usefulness and inten- act as motivators,making individuals more inclined to use tion to cryptocurrency use) is ranked using ANN sensitivity new technology; however, insecurity and discomfort act as analysis to confirm the PLS-SEM results. The findings of the inhibitors to acceptance and adoption of a given technology. ANN model generally verify the results obtained by SEM. To extend the PLS-SEM analysis, we also per- However, there are some minimal variances, which is due to formed Importance-Performance Map Analysis (IPMA) to the higher prediction accuracy and non-linear nature of ANN report additional findings and conclusions for managerial VOLUME 8, 2020 13145  O. Sohaib et al.: PLS-SEM Neural Network Approach for Understanding Cryptocurrency Adoption FIGURE 7. Importance-performance map analysis (Output: PEOU). FIGURE 8. Importance-performance map analysis (Output: PU). actions. The IPMA results are drawn on two dimensions perceived usefulness, and offer major improvement in terms (i.e., performance and importance), which is specifically of the performance level. important in order to prioritize managerial actions ,. Table 15 summarises the relative importance of the five Performance is measured on a scale from 0 to 100. Undertak- predictors (optimism, innovativeness, discomfort, insecurity ing an IPMA in our PLS path model includes determining a and perceived ease of use) of perceived usefulness. However, target construct, such as percieved ease of use. optimism is ranked higher than innovativeness in the ANN Figure 7 shows optimism is highly relevant for increasing analysis. perceived ease of use due to its strong influence. However, this factor already has a high effect (importance). Hence, TABLE 15. Summary of relative importance ranking (output: PU). there is somewhat minimal potential for a further increase. The situation is similar with the performance of the innova- tiveness factor, although the overall effects are significantly lower than optimism. The ANN ranking normalized relative importance) also confirms optimism is the most significant predictor of perceived ease of us, followed by innovative- ness. Managerial efforts should be directed at maintaining or expanding the optimism and innovativeness performance level. Similarly, Figure 7 shows that discomfort followed Finally, regarding the direct influence on the intention to by insecurity is of less importance in relation to increasing use cryptocurrency (see Figure 9 and Table 16), perceived perceived ease of use, as they have a relatively low influence. usefulness score the highest and perceived ease of use offer However, the ANN ranking placed discomfort lower than potential improvement in terms of the total effects. insecurity. Therefore, managerial actions should specifically consider addressing feelings of discomfort and insecurity to enhance perceived ease of use towards cryptocurrency use. Table 14 summarises the relative importance of the four predictors (optimism, innovativeness, discomfort, insecurity) of perceived ease of use. TABLE 14. Summary of relative importance ranking (output: PEOU). FIGURE 9. Importance-performance map analysis (Output: CUSE). The IPMA results are aligned with the structural model results; this is because both the PLS and IPMA assume linear relationships. However, ANN is capable of modelling Furthermore, in terms of perceived usefulness, Figure 8 complex linear and non-linear relationships and produces shows perceived ease of use is highly relevant for increasing more precise predictions compared to SEM methods. perceived usefulness due to its strong influence. However, The findings of this study may be useful for future cryp- amongst the technology readiness factors, discomfort and tocurrency adopters, investors and organisations. From a insecurity is of less importance in relation to increasing theoretical point of view, there is scant research on cryp- 13146 VOLUME 8, 2020 O. Sohaib et al.: PLS-SEM Neural Network Approach for Understanding Cryptocurrency Adoption TABLE 16. Summary of relative importance ranking (output: CUSE). [INNOVATIVENESS] INN1: Other people come to me for advice on digital currency such as Bitcoin. INN2: In general, I am among the first in my circle of friends to acquire a new digital currency when it appears. INN3: I can usually figure out new digital currencies with- out help from others. INN4: I keep up with the latest technological developments tocurrency adoption. This study makes a significant contri- in my areas of interest, such as digital currencies. bution to the existing literature by investigating the effect of [DISCOMFORT] technology readiness on cryptocurrency adoption. The main DIS1: If I get technical support from digital currency contribution of this study is the use of two-step PLS-SEM providers or exchanges, I will feel as if I am being taken and ANN approach provides two benefits. First, the use of advantage of by someone who knows more than I do. ANN further verifies the results obtained by the PLS-SEM DIS2: Technical support lines are not helpful because they analysis. Second,ANN is capable of modelling complex lin- do not explain things for digital currency use in a way that I ear and non-linear relationships with high predictive accuracy understand. compared to SEM methods. DIS3: Sometimes, I think that digital currency, such as In conclusion, the two-step PLS-SEM and ANN better Bitcoin, is not designed for use by ordinary people. in-depth results compared to single-step SEM method. Also, DIS4: There is no such thing as a manual for digital cur- an IPMA of the PLS-SEM findings provides a more specific rency such as Bitcoin that’s written in plain language. understanding of each facto’s importance and performance. [INSECURITY] The use of ANN enables further verification of the outcomes INS1: People are too dependent on digital currency such as obtained by the PLS-SEM analysis. Bitcoin to do things for them. INS2: Too many digital currencies distract people to the A. LIMITATIONS AND FUTURE RESEARCH point of being harmful. This study has several limitations. First, the data were col- INS3: A digital currency such as Bitcoin lowers the quality lected in one country, Australia, which may make our results of relationships by reducing personal interaction. less generalizable. Future research could consider carrying INS4: I do not feel confident doing business with digital out a cross-country comparative study with a larger data set. currency such as Bitcoin. Secondly, this study assumes digital currencies will be a form [PERCEIVED EASE OF USE] of payment in the future, and hence considers intention to use PEOU1: Learning to use the digital currency such as Bit- a digital currency, such as bitcoin. Westhuizen investi- coin would be easy for me. gated the legal status and regulation of future digital money PEOU2: Usage of the digital currency such as Bitcoin is in Australia. Thirdly, it would be interesting to include control clear and understandable to me. variables such as age and gender and compare the results. PEOU3: Overall, I find digital currency such as Bitcoin Fourthly, other factors and models such as Technology Orga- easy to use. nization Environment (TOE) may be used to examine the [PERCEIVED USEFULNESS] influence of various factors on cryptocurrency adoption and PU1: The use of digital currency such as Bitcoin enables use. me to transact online more quickly. PU2: The use of digital currency such as Bitcoin increases APPENDIX my productively. THE MEASURES PU3: The use of digital currency such as Bitcoin in my Note: These technology readiness questions were modified daily life is very useful. from the Technology Readiness Index 2.0, which is copy- [Intention to Use Cryptocurrency] righted by A. Parasuraman and Rockbridge Associates, Inc., CUSE1: I intend to use a digital currency such as Bitcoin 2014.. This scale may be duplicated only with written when it becomes widely available. permission from the original authors. CUSE2: Whenever possible, I intend to frequently use a [OPTIMISM] digital currency such as Bitcoin in my daily life. OPT1: New digital currencies such as Bitcoin contribute to CUSE3: I intend to use a digital currency when it is a better quality of life. legally accepted as a form of payment in the country of my OPT2: A digital currency such as Bitcoin gives me more residence. freedom of mobility. OPT3: A digital currency such as Bitcoin gives people REFERENCES more control over their daily lives. D. Tapscott and A. Tapscott, Blockchain Revolution: How the Technology OPT4: A digital currency such as Bitcoin makes me more Behind Bitcoin and Other Cryptocurrencies is Changing the World. Irvine, productive in my personal life. CA, USA: Portfolio, 2016, p. 368. VOLUME 8, 2020 13147  O. Sohaib et al.: PLS-SEM Neural Network Approach for Understanding Cryptocurrency Adoption G. O. Karame, E. Androulaki, M. Roeschlin, A. Gervais, and S. Čapkun, F. Glaser, ‘‘Pervasive decentralisation of digital infrastructures: A frame- ‘‘Misbehavior in bitcoin: A study of double-spending and accountability,’’ work for blockchain enabled system and use case analysis,’’ in Proc. J. ACM Trans. Inf. Syst. Secur., vol. 18, no. 1, pp. 1–32, 2015. 50th Hawaii Int. Conf. Syst. Sci. (HICSS), Waikoloa Village, Jan. 2017, E.-T. Cheah and J. Fry, ‘‘Speculative bubbles in bitcoin markets? An empir- pp. 1543–1552. ical investigation into the fundamental value of Bitcoin,’’ Econ. Lett., M. Risius and K. Spohrer, ‘‘A blockchain research framework,’’ Bus. Inf. vol. 130, pp. 32–36, May 2015. Syst. Eng., vol. 59, no. 6, pp. 385–409, Dec. 2017. M. Polasik, A. I. Piotrowska, T. P. Wisniewski, R. Kotkowski, and D. Efanov and P. Roschin, ‘‘The all-pervasiveness of the blockchain tech- G. Lightfoot, ‘‘Price fluctuations and the use of bitcoin: An empirical nology,’’ Procedia Comput. Sci., vol. 123, pp. 116–121, Jan. 2018. inquiry,’’ Int. J. Electron. Commerce, vol. 20, no. 1, pp. 9–49, Sep. 2015. C. Tucker, ‘‘Identifying formal and informal influence in technology J. Bouoiyour and R. Selmi, ‘‘What does bitcoin look like?’’ Ann. Econ. adoption with network externalities,’’ Manage. Sci., vol. 54, no. 12, Finance, vol. 16, no. 2, pp. 449–492, 2015. pp. 2024–2038, 2008. L. Kristoufek, ‘‘BitCoin meets Google trends and Wikipedia: Quantifying V. Venkatesh, M. G. Morris, B. Gordon, and F. D. Davis, ‘‘User acceptance the relationship between phenomena of the Internet era,’’ Sci. Rep., vol. 3, of information technology: Toward a unified view,’’ MIS Quart., vol. 27, p. 3415, Apr. 2013. no. 3, pp. 425–478, Sep. 2003. D. Garcia, C. J. Tessone, P. Mavrodiev, and N. Perony, ‘‘The digital traces C. Decker and R. Wattenhofer, ‘‘Information propagation in the bitcoin of bubbles: Feedback cycles between socio-economic signals in the Bitcoin network,’’ in Proc. IEEE P2P, Sep. 2013, pp. 1–10. economy,’’ J. Roy. Soc. Interface, vol. 11, no. 99, 2014, Art. no. 20140623. C. Tsanidis, D.-M. Nerantzaki, G. Karavasilis, V. Vrana, and R. Böhme, N. Christin, B. Edelman, and T. Moore, ‘‘Bitcoin: Eco- D. Paschaloudis, ‘‘Greek consumers and the use of Bitcoin,’’ Bus. nomics, technology, and governance,’’ J. Econ. Perspect., vol. 29, no. 2, Manage. Rev., vol. 6, no. 2, pp. 295–302, 2015. pp. 213–238, 2015. X. Li and C. A. Wang, ‘‘The technology and economic determinants of J. Bohr and M. Bashir, ‘‘Who uses bitcoin? An exploration of the bitcoin cryptocurrency exchange rates: The case of Bitcoin,’’ Decis. Support Syst., community,’’ in Proc. 12th Annu. Int. Conf. Privacy, Secur. Trust, 2014, vol. 95, pp. 49–60, Mar. 2017. pp. 94–101. F. Hawlitschek, B. Notheisen, and T. Teubner, ‘‘The limits of trust-free J. Silinskyte, ‘‘Understanding bitcoin adoption: Unified theory of accep- systems: A literature review on blockchain technology and trust in the tance and use of technology (UTAUT) application,’’ M.S. thesis, Leiden sharing economy,’’ Electron. Commerce Res. Appl., vol. 29, pp. 50–63, Inst. Adv. Comput. Sci., Leiden, The Netherlands, 2014. May/Jun. 2018. J. Wu and A. Lederer, ‘‘A meta-analysis of the role of environment-based D. Lee Kuo Chuen, L. Guo, and Y. Wang, ‘‘Cryptocurrency: A new invest- voluntariness in information technology acceptance,’’ MIS Quart., vol. 33, ment opportunity?’’ J. Alternative Investments, vol. 20, no. 3, pp. 16–40, no. 2, pp. 419–432, 2009. 2017. P. Godoe and T. S. Johansen, ‘‘Understanding adoption of new technolo- CoinMarketCap. (May 2019). Bitcoin. [Online]. Available: https:// gies: Technology readiness and technology acceptance as an integrated coinmarketcap.com/currencies/bitcoin/#charts concept,’’ J. Eur. Psychol. Students, vol. 3, no. 1, pp. 38–52, 2012. J. K. Darlington, ‘‘The future of bitcoin: Mapping the global adoption M. Mcdougall, ‘‘An investigation of the theory of disruptive innovation: of world’s largest cryptocurrency through benefit analysis,’’ Univ. Ten- Does the cryptocurrency bitcoin have the potential to be a disruptive nessee, Knoxville, Tennessee, Tech. Rep., May 2014. [Online]. Available: innovation relative to an existing market?’’ M.S. thesis, Edinburgh Napier https://trace.tennessee.edu/utk_chanhonoproj/1770 Univ., Edinburgh, Scotland, 2014. S. Y. Lam, J. Chiang, and A. Parasuraman, ‘‘The effects of the dimensions A. Parasuraman and C. L. Colby, ‘‘An updated and streamlined technology of technology readiness on technology acceptance: An empirical analysis,’’ readiness index: TRI 2.0,’’ J. Service Res., vol. 18, no. 1, pp. 59–74, J. Interact. Marketing, vol. 22, no. 4, pp. 19–39, Nov. 2008. Feb. 2015. A. Parasuraman, ‘‘Technology readiness index (Tri): A multiple-item scale M. Ratchford and M. Barnhart, ‘‘Development and validation of the tech- to measure readiness to embrace new technologies,’’ J. Service Res., vol. 2, nology adoption propensity (TAP) index,’’ J. Bus. Res., vol. 65, no. 8, no. 4, pp. 307–320, May 2000. pp. 1209–1215, Aug. 2012. C.-H. Lin, H.-Y. Shih, and P. J. Sher, ‘‘Integrating technology readiness into M. Son and K. Han, ‘‘Beyond the technology adoption: Technology readi- technology acceptance: The TRAM model,’’ Psychol. Marketing, vol. 24, ness effects on post-adoption behavior,’’ J. Bus. Res., vol. 64, no. 11, no. 7, pp. 641–657, 2007. pp. 1178–1182, Nov. 2011. F. D. Davis, ‘‘Perceived usefulness, perceived ease of use, and user accep- N. Larasati, Widyawan, and P. I. Santosa, ‘‘Technology readiness and tance of information technology,’’ MIS Quart., vol. 13, no. 3, pp. 319–340, technology acceptance model in new technology implementation process 1989. in low technology SMEs,’’ Int. J. Innov., Manage. Technol., vol. 8, no. 2, J. S. C. Lin and H. C. Chang, ‘‘The role of technology readiness in self- pp. 113–117, 2017. service technology acceptance,’’ Manag. Service Qual., Int. J., vol. 21, K. Koivisto, M. Makkonen, L. Frank, and J. Riekkinen, ‘‘Extending the no. 4, pp. 424–444, 2011. technology acceptance model with personal innovativeness and technology R. Walczuch, J. Lemmink, and S. Streukens, ‘‘The effect of service readiness: A comparison of three models,’’ in Proc. BLED, 2016. employees’ technology readiness on technology acceptance,’’ Inf. Man- H. Hallikainen and T. Laukkanen, ‘‘How technology readiness explains age., vol. 44, no. 2, pp. 206–215, Mar. 2007. acceptance and satisfaction of digital services in B2B healthcare sector?’’ P. Basgoze, ‘‘Integration of technology readiness (TR) into the technology in Proc. PACIS. Chiayi, Taiwan: AIS, 2016, pp. 294–306. acceptance model (TAM) for m-shopping,’’ Int. J. Sci. Res. Innov. Technol., vol. 2, no. 3, pp. 26–35, 2015. A. J. Connolly and A. Kick, ‘‘What differentiates early organization S. S. Zabukovšek, Z. Kalinic, S. Bobek, and P. Tominc, ‘‘SEM–ANN based adopters of bitcoin from non-adopters?’’ presented at the 21st Amer. Conf. research of factors’ impact on extended use of ERP systems,’’ Central Eur. Inf. Syst., Puerto Rico, Fajardo, 2015. J. Oper. Res., vol. 27, pp. 703–735, Nov. 2018. E. Karahanna, D. Straub, and N. L. Chervany, ‘‘Information technology P.-Y. Foo, V.-H. Lee, G. W.-H. Tan, and K.-B. Ooi, ‘‘A gateway to realising adoption across time: A cross-sectional comparison of pre-adoption and sustainability performance via green supply chain management practices: post-adoption beliefs,’’ MIS Quart., vol. 23, no. 2, pp. 183–213, 1999. A PLS-ANN approach,’’ Expert Syst. Appl., vol. 107, pp. 1–14, Oct. 2018. K. M. Kuo, C. F. Liu, and C. C. Ma, ‘‘An investigation of the effect A. Y.-L. Chong, M. J. Liu, J. Luo, and O. Keng-Boon, ‘‘Predicting RFID of nurses’ technology readiness on the acceptance of mobile electronic adoption in healthcare supply chain from the perspectives of users,’’ Int. medical record systems,’’ BMC Med Inf. Decis. Making, vol. 13, p. 88, J. Prod. Econ., vol. 159, pp. 66–75, Jan. 2015. Aug. 2013. J. Henseler, T. K. Dijkstra, M. Sarstedt, C. M. Ringle, A. Diamantopoulos, P. Upadhyay and M. Chattopadhyay, ‘‘Examining mobile based payment D. W. Straub, D. J. Ketchen, J. F. Hair, G. T. M. Hult, and J. R. Cantalone, services adoption issues: A new approach using hierarchical clustering ‘‘Common beliefs and reality about PLS: Comments on Rönkkö and Ever- and self-organizing maps,’’ J. Enterprise Inf. Manage., vol. 28, no. 4, mann (2013),’’ Organizational Res. Methods, vol. 17, no. 2, pp. 182–209, pp. 490–507, Jul. 2015. 2014. R. Buyle, M. Van Compernolle, E. Vlassenroot, Z. Vanlishout, P. Mechant, J. F. Hair, L. M. Matthews, R. L. Matthews, and M. Sarstedt, ‘‘PLS- and E. Mannens, ‘‘‘Technology readiness and acceptance model’ as a SEM or CB-SEM: Updated guidelines on which method to use,’’ Int. predictor for the use intention of data standards in smart cities,’’ Media J. Multivariate Data Anal., vol. 1, no. 2, pp. 107–123, 2017. Commun., vol. 6, no. 4, pp. 127–139, 2018. 13148 VOLUME 8, 2020 O. Sohaib et al.: PLS-SEM Neural Network Approach for Understanding Cryptocurrency Adoption F. Liébana-Cabanillas, V. Marinković, and Z. Kalinić, ‘‘A SEM-neural J. F. Hair, M. Sarstedt, C. M. Ringle, and S. P. Gudergan, Advanced Issues network approach for predicting antecedents of m-commerce acceptance,’’ in Partial Least Squares Structural Equation Modeling. Newbury Park, Int. J. Inf. Manage., vol. 37, no. 2, pp. 14–24, 2017. CA, USA: Sage, 2017. C. M. Ringle and R. R. Sinkovics, ‘‘The use of partial least squares C van der Westhuizen, ‘‘Future digital money: The legal status and regula- path modeling in international marketing,’’ Adv. Int. Marketing, vol. 20, tion of bitcoin in Australia,’’ M.S. thesis, School Law, Univ. Notre Dame pp. 277–319, Mar. 2009. Australia, Fremantle, WA, Australia, 2017. K.-B. Ooi and G. W.-H. Tan, ‘‘Mobile technology acceptance model: An investigation using mobile users to explore smartphone credit card,’’ Expert Syst. Appl., vol. 59, pp. 33–46, Oct. 2016. J.-J. Hew, M. N. B. A. Badaruddin, and M. K. Moorthy, ‘‘Crafting a smartphone repurchase decision making process: Do brand attachment and gender matter?’’ Telematics Informat., vol. 34, no. 4, pp. 34–56, Jul. 2017. C. M. Ringle, S. Wende, and J.-M. Becker. (2014). Smartpls 3. Hamburg: OSAMA SOHAIB is currently a Lecturer with SmartPLS. [Online]. Available: http://www.smartpls.com the School of Information, Systems and Mod- W. Reinartz, M. Haenlein, and J. Henseler, ‘‘An empirical comparison of eling, University of Technology Sydney (UTS). the efficacy of covariance-based and variance-based SEM,’’ Int. J. Res. His work has published in various reputable Marketing, vol. 26, no. 4, pp. 332–344, 2009. journals such as Computers & Industrial Engi- F. Hair, J. S. Marko, H. Lucas, and G. K. Volker, ‘‘Partial least squares neering, IEEE ACCESS, Mobile Networks structural equation modeling (PLS-SEM) an emerging tool in business and Applications, the International Journal research,’’ Eur. Bus. Rev., vol. 26, no. 2, pp. 106–121, 2014. of Disaster Risk Reduction, the Journal of Ambient J. F. Hair, C. M. Ringle, and M. Sarstedt, ‘‘PLS-SEM: Indeed a silver Intelligence and Humanized Computing, the Jour- bullet,’’ J. Marketing Theory Pract., vol. 19, no. 2, pp. 139–151, 2011. nal of Global Information Management, and Sus- M. Rönkkö and J. Evermann, ‘‘A critical examination of common beliefs tainability. His research interests include decision-making, e-services, HCI, about partial least squares path modeling,’’ Organizational Res. Methods, and survey methods. vol. 16, no. 3, pp. 425–448, 2013. M. Sarstedt, J. F. Hair, C. M. Ringle, K. O. Thiele, and S. P. Gudergan, ‘‘Estimation issues with PLS and CBSEM: Where the bias lies!’’ J. Bus. Res., vol. 69, no. 10, pp. 3998–4010, Oct. 2016. N. Kock, ‘‘Common method bias in PLS-SEM: A full collinearity assess- ment approach,’’ Int. J. e-Collaboration, vol. 11, no. 4, pp. 1–10, 2015. J. Henseler and M. Sarstedt, ‘‘Goodness-of-fit indices for partial least WALAYAT HUSSAIN received the Ph.D. degree squares path modeling,’’ Comput. Statist., vol. 28, no. 2, pp. 565–580, from the University of Technology Sydney. Apr. 2013. He worked as a Lecturer and an Assistant J. Hair, C. L. Hollingsworth, A. B. Randolph, and A. Y. L. Chong, Professor at BUITEMS for many years. He is ‘‘An updated and expanded assessment of PLS-SEM in information sys- currently a Lecturer with the Faculty of Engi- tems research,’’ Ind. Manag. Data Syst., vol. 117, no. 3, pp. 442–458, 2017. neering and IT, University of Technology Sydney, S. Petter, ‘‘‘Haters gonna hate’: PLS and information systems research,’’ Australia. He published in various top-ranked rep- SIGMIS Database, vol. 49, no. 2, pp. 10–13, 2018. utable journals and conferences such as the Com- C. Fornell and D. F. Larcker, ‘‘Evaluating structural equation models with puter Journal, Information Systems, IEEE ACCESS, unobservable variables and measurement error,’’ J. Marketing Res., vol. 18, Future Generation Computer Systems, Computers no. 1, pp. 39–50, 1981. & Industrial Engineering, Mobile Networks and Applications, the Journal of J. Henseler, C. M. Ringle, and M. Sarstedt, ‘‘A new criterion for assessing Ambient Intelligence and Humanized Computing, FUZZ-IEEE, and ICONIP. discriminant validity in variance-based structural equation modeling,’’ His research areas are business intelligence, cloud computing, and usability J. Acad. Marketing Sci., vol. 43, no. 1, pp. 115–135, 2015. engineering by focusing on providing an informed decision to different stake- G. Shmueli, S. Ray, J. M. V. Estrada, and S. B. Chatla, ‘‘The elephant in holders. He received three international and one national research awards the room: Predictive performance of PLS models,’’ J. Bus. Res., vol. 69, and recognitions till date from his research. He was a recipient of 2016 FEIT no. 10, pp. 4552–4564, Oct. 2016. HDR Publication Award by the University of Technology Sydney, Australia. P. N. Sharma, G. Shmueli, M. Sarstedt, N. Danks, and S. Ray, ‘‘Prediction- oriented model selection in partial least squares path modeling,’’ Decis. Sci., to be published. J. Evermann and M. Tate, ‘‘Assessing the predictive performance of structural equation model estimators,’’ J. Bus. Res., vol. 69, no. 10, pp. 4565–4582, Oct. 2016. L.-Y. Leong, T.-S. Hew, V.-H. Lee, and K.-B. Ooi, ‘‘An SEM-artificial- MUHAMMAD ASIF received the M.S. and Ph.D. neural-network analysis of the relationships between SERVPERF, cus- degrees from AIT, in 2009 and 2012, respectively, tomer satisfaction and loyalty among low-cost and full-service airline,’’ on HEC Foreign Scholarship. During the course Expert Syst. Appl., vol. 42, no. 19, pp. 6620–6634, Nov. 2015. of time, he was a Visiting Researcher with the L.-Y. Leong, T.-S. Hew, G. W.-H. Tan, and K.-B. Ooi, ‘‘Predicting the National Institute of Information, Tokyo, Japan. determinants of the NFC-enabled mobile credit card acceptance: A neural He was a Research Scholar with the Computer Sci- networks approach,’’ Expert Syst. Appl., vol. 40, no. 14, pp. 5604–5620, Oct. 2013. ence and Information Management Department, Asian Institute of Technology, Thailand. He is F. T. S. Chan and A. Y. L. Chong, ‘‘A SEM-neural network approach for understanding determinants of interorganizational system standard adop- currently a Chairman with the Department of tion and performances,’’ Decis. Support Syst., vol. 54, no. 1, pp. 621–630, Computer Science, National Textile University, Dec. 2012. Faisalabad. He has worked on some projects including the Air Traffic Control C. M. Ringle and M. Sarstedt, ‘‘Gain more insight from your PLS-SEM System of Pakistan Air force. He is also a Permanent Member of Punjab Pub- results: The importance-performance map analysis,’’ Ind. Manage. Data lic Service Commission (PPSC) as an Advisor and a Program Evaluator at the Syst., vol. 116, no. 9, pp. 1865–1886, Oct. 2016. National Computing Education Accreditation Council (NCEAC) Islamabad. C. Hock, C. M. Ringle, and M. Sarstedt, ‘‘Management of multi-purpose He is also serving as an Associate Editor of IEEE ACCESS, the prestigious stadiums: Importance and performance measurement of service inter- journal of IEEE. He is serving as a Reviewer for a number of reputed journals faces,’’ Int. J. Services Technol. Manage., vol. 14, nos. 2–3, pp. 188–207, and also authored a number of research articles in reputed journals and Jan. 2010. conferences. VOLUME 8, 2020 13149  O. Sohaib et al.: PLS-SEM Neural Network Approach for Understanding Cryptocurrency Adoption MUHAMMAD AHMAD is currently an Assistant MANUEL MAZZARA received the Ph.D. degree Professor with the Department of Computer Engi- in computing science from the University of neering, Khwaja Freed University of Engineering Bologna. He is currently a Professor of com- and Information Technology. He has published puter science with the Institute of Software dozens of articles in Top tier journals/conferences. Development and Engineering, Innopolis Univer- His current research interests include machine sity, Innopolis, Russia. He has published 100 of learning, computer vision, remote sensing, hyper- articles in Top tier journals/conferences. His spectral imaging, and wearable computing. He is research interests include software engineering, a regular reviewer for several top tier journals service-oriented architecture and programming, including but not limited to Nature, IEEE TIE, concurrency theory, formal methods, software ver- IEEE TNNLS, IEEE TGRS, IEEE TIP, IEEE GRSL, IEEE GRSM, ification, and artificial intelligence. IEEE JSTAR, the IEEE TRANSACTIONS ON MOBILE COMPUTING, the IEEE TRANSACTIONS ON MULTIMEDIA, the IEEE TRANSACTIONS ON INDUSTRIAL APPLICATIONS, Remote Sensing, IEEE ACCESS, the IEEE COMPUTERS, the IEEE SENSORS, NCAA, Measurement Science and Technology, IET Image Process- ing, Transactions on Internet and Information Systems, and many more. 13150 VOLUME 8, 2020