Computers in Human Behavior 90 (2019) 181–187 Contents lists available at ScienceDirect Computers in Human Behavior journal homepage: www.elsevier.com/locate/comphumbeh Full length article A hybrid modeling approach for predicting the educational use of mobile cloud computing services in higher education T Ibrahim Arpaci Tokat Gaziosmanpasa University, Department of Computer Education and Instructional Technology, Tokat, Turkey A R T I C LE I N FO A B S T R A C T Keywords: Information management Cloud computing Artificial intelligence Machine learning The decision to integrate mobile cloud computing (MCC) in education without determining optimal use scenarios is a universal problem as the adoption of such services becomes widespread. Accordingly, this study developed and validated a predictive model that explains the role of students' information management (i.e. retrieve, store, share, and apply) practices in predicting their attitudes toward using the MCC services for educational purposes. This study validated the model by the complementary use of machine learning algorithms alongside a classical SEM-based approach based on data collected from 308 undergraduate students. The SEM results indicated that the students' information management (i.e. retrieve, store, share, and apply) practices were significantly associated with their attitudes, which were significantly associated with the behavioral intentions. The structural model explained a significant portion of the variance in the behavioral intentions. Likewise, the classifier model suggested that the students’ information management practices and attitudes predicted the behavioral intentions. Further, the applied algorithms predicted the behavioral intentions with an accuracy of more than 72% in most cases. Thereby, the study extended an original theory (TRA) into the MCC area by using a multi-analytical approach. The findings implied that employing the MCC services for personal information management should be supported and encouraged in the higher education by designing authentic learning environments and scaffolding the students in using such services. 1. Introduction Mobile cloud computing (MCC) as a new distributed computing paradigm can be defined as an infrastructure, application or process, where the data storage and processing migrated from smart mobile technologies to the distributed cloud servers (Dinh, Lee, Niyato, & Wang, 2013). The MCC has gained a substantial attention of both the organizations and individuals as a promising solution for the ubiquitous environments, in which data storage and processing occur over a “cloud” via the Internet (Park & Kim, 2014). Investing in the MCC, organizations improved their capacity and capabilities without the cost of installing new infrastructure or software (Subashini & Kavitha, 2011). Further, the MCC reduced the initial installation and maintenance costs, and thereby, promoted the efficiency and green technology (Aepona, 2018). There are various cloud deployment-models such as private, public, hybrid, and community cloud (Shon, Cho, Han, & Choi, 2014). The study focused on the public cloud that allocates resources on a per-user basis through applications such as Dropbox, SkyDrive, and Google Drive. However, the MCC service models for individuals include “platform as a service” (PaaS) for automatic content synchronizing and “infrastructure as a service” (IaaS) for information storage and management. On the other hand, these services provide enterprises with the IaaS for an enterprise infrastructure building and “software as a service” (SaaS) for an enterprise solution provider (Arpaci, 2016). The MCC provides users at the organizational and individual level with several advantages and given these advantages the MCC services are considered to be the growth engine of the industry 4.0 (Park & Kim, 2014). For example, the MCC increases data storage capacity and allows an efficient data synchronization and information management in a ubiquitous environment (Park & Kim, 2014). Besides, energy efficiency is a critical issue since battery power of the mobile technologies is limited. Migration of the complex processing from mobile technologies to the cloud servers extends the battery life (Cuervo et al., 2010). Furthermore, service providers have certain security mechanisms and backup systems to protect user data. Therefore, saving documents and files on the distributed cloud servers may enhance the reliability by reducing virus-like threats and data loss (Arpaci, 2016). Smart mobile technologies change the way of information is managed (Ogiela, 2017). For example, using mobile Internet on smartphones provides students with ubiquitous access to information and learning materials. Further, social networking and communication E-mail address: ibrahim.arpaci@gop.edu.tr. https://doi.org/10.1016/j.chb.2018.09.005 Received 20 January 2018; Received in revised form 16 August 2018; Accepted 10 September 2018 Available online 11 September 2018 0747-5632/ © 2018 Elsevier Ltd. All rights reserved. Computers in Human Behavior 90 (2019) 181–187 I. Arpaci 3. Theoretical background applications allow them to share information with peers. More importantly, the integrated technologies enable them to use information stored in the MCC services in problem solving and decision-making. Considering the extensive use of mobile devices such as smartphones, tablets, and smartwatches by undergraduate students, it is important to find creative ways for an effective integration of these technologies into the higher education. Such an integration may enable students to learn by using the technologies, which they are familiar and confident with, and highly motivated to use. It is important to note that providing an effective management of personal information through information retrieval, access, storage, sharing, and application is one of the main advantages of the MCC services. However, the decision to integrate the MCC in educational settings without determining optimal use scenarios is a universal problem as the adoption of such services becomes widespread. Thus, this study investigated adoption of the MCC among undergraduate students by focusing on the use of these services for personal information management. Most of the studies reviewed applied or extended a theory such as TAM (Davis, 1989) or UTAUT (Venkatesh, Morris, Davis, & Davis, 2003) by conducting an explanatory statistical analysis. However, such an approach has two main limitations. First, each theory was developed and tested for a specific domain, and therefore, may not work well in other domains or areas. Thus, this study extended the original “Theory of Reasoned Action” (TRA) with additional constructs to better explain the research context. Second, previous studies pointed out statistical limitations of the technology adoption studies and emphasized the importance of employing a multi-analytical approach by combining predictive analytics and causal explanatory statistical analysis to validate a predictive model (Sharma et al., 2016; Sharma, Joshi, & Sharma, 2016; Tan, Ooi, Leong, & Lin, 2014). The study therefore tested the research model by using both the classical SEM approach and artificial intelligence techniques (i.e. machine learning). The study employed the SEM approach to understand causal relationships and the (complementary) study applied the machine learning algorithms to predict the behavioral intentions based on the proposed constructs. Further, this study approaches the MCC adoption from an information management perspective. The fact that providing an effective information management was one of the main advantages of the MCC services supports the idea that individuals' information management practices may have a critical role in their adoption decision. Accordingly, this study investigated the role of students’ information management practices on their attitudes toward using the MCC services in educational settings. The TRA postulated that beliefs affect attitudes, which positively influence behavioral intentions, whereas the intentions ultimately influence the actual behavior (Fishbein & Ajzen, 1975). Benbasat and Barki (2007) suggested using the original theory (TRA) to provide a strong theoretical grounding for a novel model by incorporating different antecedents relevant to the nature of ICT integration and use in diverse settings. Likewise, Nistor (2014) argued that “educational technology acceptance” studies should consider characteristics of the educational context. Considering the shortcomings of the technology acceptance models and analysis methods suggested by (Benbasat & Barki, 2007), this study identified the key antecedents of the attitudes towards using the MCC services and employed an innovative approach comprising both machine learning algorithms and the SEM, and thereby, strengthened the study design. Fig. 1 illustrates the research model that suggests the behavioral intentions (continued use intentions) are predicted by the attitudes, which are predicted by the students’ information management practices such as information retrieval, storage, sharing, and application. 2. Literature review The MCC is a recent research area and there are numerous studies on the adoption of the MCC at the both organizational and individual level. For example, Ul Amin, Inayat, Shahzad, Saleem, and Aijun (2017) applied the “Unified Theory of Acceptance and Use of Technology” (UTAUT) to identify key predictors of the MCC adoption by healthcare professionals. Their results suggested the UTAUT constructs (i.e. performance-expectancy, social influence, effort-expectancy) were significant in determining the behavioral intentions. In another study, Lian, Yen, and Wang (2014) investigated the factors affecting the cloud computing adoption in Taiwan's hospital industry based on the HOT-fit (Human-Organization-Technology fit) model and TOE (TechnologyOrganization-Environment) framework. The results suggested that technical competence, data security, cost, complexity, and top management support were critical factors in the adoption. Chen, Chen, and Lee (2018) investigated key factors affecting the organizations' adoption of the cloud services. Their results suggested that top-managementsupport has a vital role in the adoption at the organizational level. Lee (2016) identified the key factors explaining the adoption of cloud services at the individual level based on the “Diffusion of Innovations” (DOI) theory. The results suggested the relative advantage, observability, and self-efficacy were significant antecedents of the adoption. Park and Kim (2014) focused on key factors predicting the individuals' acceptance of the MCC services. Results indicated that connectedness, perceived mobility, service quality, security, and satisfaction were significant predictors of the user acceptance. Wang and Huang (2016) reported that social influence was a key predictor of the students’ intention to use cloud services. Further, they reported that scaffolding in problem-solving and training were useful in familiarizing the students with such services. Arpaci (2016) proposed a theoretical structural model based on the “Technology Acceptance Model” (TAM) and suggested that subjective norm, trust, and perceived-usefulness play significant roles in predicting the students' attitudes towards using the MCC services. In another study, Sharma, Al-Badi, Govindaluri, and Al-Kharusi (2016) developed a hybrid model based on the TAM to predict factors affecting the adoption of the cloud computing by IT professionals. They employed linear regression along with neural network analysis and found that job opportunity was the most significant predictor of the adoption. Shiau and Chau (2016) developed a theoretical structural model based on the TAM and motivational model to predict behavioral intentions to join a cloud computing class. The results suggested subjective norms and attitudes were positively associated with the behavioral intention. Kim and Kim (2018) proposed a research model in order to identify the antecedents of the MCC adoption. The results suggested that trust, convenience, and perceived uncertainty were significant factors in the individuals’ adoption decision. Information Retrieval Information Storage Attitude Information Sharing .77** R2 = .33, e = .17 Behavioral Intention R2 = .59, e = .17 Information Application *p< .01, **p< .001, Chi-Square= 393.25, DF= 235, Chi-Square/DF=1.67 Fig. 1. The structural model. 182 Computers in Human Behavior 90 (2019) 181–187 I. Arpaci Rogers's (2003, p.19) DOI theory suggested that the stages by which an individual adopts a new technology or innovation, whereby diffusion is accomplished include “awareness of the need for an innovation, decision to adopt (or reject) the innovation, initial use of the innovation to test it, and continued use of the innovation.” Accordingly, a training program introduced the participants to personal information management, which refers to the practice of activities an individual performs to create or acquire, store, retrieve, share, and apply the information required to complete the tasks. The participants employed the mobile web services (i.e. search, maps), mobile cloud applications (i.e. Google Drive, iCloud, Dropbox) and social media applications to perform the training tasks. the use of such services. 3.1. Hypotheses 4.1. Sample 3.1.1. Behavioral intentions and attitudes “Behavioral intention” was defined as “the degree of an individual's belief that he or she will continue to use a system” (Arpaci, 2017, p. 384). However, attitudes towards using a technology was defined as “an individual's overall affective reaction to use a particular system” (Arpaci, 2017, p. 384). This study therefore hypothesized that the attitude toward using the MCC services would be positively associated with the behavioral intention to use such services for educational purposes. Further, the attitude would be a predictor of the behavioral intention. Therefore; A sample of 308 university students were recruited from a public university in Turkey. The study included 186 females (60.4%) and 122 (39.6%) males with a mean age of 21.88 years (SD = 2.37). Majority of the students (42.5%) were seniors; 12.3% were freshmen, 27.6% were sophomore, and 17.5% were juniors. H4b. Information sharing would predict the behavioral intention to use the MCC services for educational purposes. H5a. An increase in the volume of information that's applied via the use of the MCC services is positively associated with the attitude towards the use of such services. H5b. Information application would predict the behavioral intention to use the MCC services for educational purposes. 4. Method 4.2. Procedure IRB of the affiliated university approved the research and the procedures complied with the ethical standards of the institutional board guidelines. All participants were informed about the purpose of the research after an informed consent was obtained. The study was conducted in regularly scheduled IT classes (4 h per week) during 14 weeks. The participants performed practical implementations of the theoretical knowledge during the interactive tasks such as: 1) Search for research articles in electronic databases (i.e. Scopus, WOS, and Google Scholar), 2) Save an annotated bibliography of the articles to the MCC services (i.e. OneDrive, Dropbox, and Google Drive), 3) Form a communities of practice and use the MCC services along with social networks to share files, communicate with others, and build a common cognition to complete the tasks. Finally, the survey instrument with explicit assurance that anonymity is guaranteed was administered during the last week of the regular classes. H1a. Attitude towards using the MCC services would be positively associated with the behavioral intention. H1b. Attitude towards using the MCC services would predict the behavioral intention. 3.1.2. Information management practices Information is processed, categorized, organized or structured data (Davenport & Prusak, 2000; Nonaka, 1994). However, knowledge is a combination of skills, information, experience, and insights (Nonaka & Takeuchi, 1995). Knowledge includes organizational routines, values, and practices; therefore, this study preferred to use the term “information management” instead of “knowledge management.” This study defined information management as practices that help retrieve and store information, share it with others, and apply it to solve problems and make better and timely decisions (Alavi & Leidner, 2001; Mitchell, 2003). Turban, Sharda, and Delen (2011) suggested the use of information technologies to enhance information management practices of retrieval and access, transfer, storage, and application. In the same vein, the present study suggested the integration of smart mobile technologies into the higher education to support personal information management. The more information managed by using the MCC services the more favorable the attitude towards using such services for educational purposes. Further, the students’ information management practices would be a predictor of the behavioral intention to use the MCC services for educational purposes. Accordingly, 4.3. Instruments The measurement instrument for the variables “behavioral intentions” and “attitudes” was a Likert scale adapted from Ajzen's theory (1991). Moreover, the measurement instrument for the information management practices of retrieval, storage, sharing, and application was adapted from a previously validated and reliable instrument (Arpaci, 2017). The final instrument has 24 items with a five-point Likert-type scale ranging from “strongly disagree” to “strongly agree.” 5. Results 5.1. Internal reliability and validity H2a. An increase in the volume of information that's retrieved via the use of the MCC services is positively associated with the attitude towards the use of such services. IBM SPSS (ver. 25) and IBM AMOS (ver. 23) were used to screen and analyze the data. Construct validity along with reliability of the scales were assessed by conducting an EFA (See Table 1). Results revealed that each item had a communality value and a factor loading greater than the critical threshold (0.40) suggested by Field (2005). Cronbach's alpha reliability analysis results indicated a good homogeneity and reliability among the items. The AVE values were greater than 0.50, indicating adequate convergent validity for all constructs (Hair, Tatham, Anderson, & Black, 2006). H2b. Information retrieval would predict the behavioral intention to use the MCC services for educational purposes. H3a. An increase in the volume of information that's stored via the use of the MCC services is positively associated with the attitude towards the use of such services. H3b. Information storage would predict the behavioral intention to use the MCC services for educational purposes. 5.2. The structural model H4a. An increase in the volume of information that's shared via the use of the MCC services is positively associated with the attitude towards Kaiser's (1970) “KMO measure of sampling adequacy” and Bartlett's 183 I. Arpaci Table 1 Instrument reliability and validity. Item α Item-total correlation Factor load Communality Total variance explained CR AVE Behavioral intention CUI1. “I intend to continue to use mobile cloud services for educational purposes in the future.” CUI2. “I plan to continue to use mobile cloud services for personal information management in the future.” CUI3. “I predict that I would frequently use mobile cloud services in the future.” CUI4. “I predict that I would frequently use mobile cloud services for educational purposes.” CUI5. “I predict that I would frequently use mobile cloud services for personal information management.” A1. “Using mobile cloud services for educational purposes is a good idea.” A2. “Using mobile cloud services to create a retrievable archive of personal information is pleasant.” A3. “Using social media applications on my smartphone to exchange information is fun.” A4. “Using mobile cloud services for personal information management is a wise idea.” A5. “Using mobile cloud services for educational purposes makes learning more interesting.” A6. “I like the idea of using mobile cloud services for personal information management.” IR1. “Using mobile Internet on my smartphone enables to retrieve learning materials and information anytime and anywhere.” IR2. “I access course materials and information using mobile cloud services on my smartphone.” IR3. “Using mobile cloud services on my smartphone enables quick access to learning materials and information resources.” IR4. “Using mobile cloud services on my smartphone enables ubiquitous access to e-databases and e-journals.” IS1. “Using mobile cloud services on my smartphone enables the retrievable storage of electronic information and documents.” IS2. “I store course materials and documents using mobile cloud services on my smartphone.” IS3. “Using mobile cloud services enables me to store learning materials and information with ubiquitous access.” ISh1. “Using social media applications on my smartphone enables me to exchange information and documents with classmates.” ISh2. “I share course materials and information with classmates using mobile cloud services on my smartphone.” ISh3. “I share learning materials and documents with classmates using mobile cloud services on my smartphone.” IA1: “I apply knowledge and experience gained by using mobile cloud services to complete learning tasks.” IA2. “I use knowledge obtained by using mobile cloud services in decision making processes.” IA3. “I employ knowledge and intelligence gained from using mobile cloud services in problem-solving activities.” .91 .76 .78 .77 .79 .79 .73 .78 .67 .75 .68 .67 .75 .85 .86 .86 .87 .87 .82 .86 .78 .83 .78 .77 .86 .72 .75 .74 .76 .76 .67 .74 .61 .69 .61 .60 .74 74.34 .91 .68 65.30 .89 .59 78.43 .91 .51 .85 .84 .92 .92 .84 .84 .91 .74 .81 .85 .91 .72 .83 84.06 .91 .52 .71 .83 .80 .51 .93 .91 .80 .86 .83 .64 64.61 .71 .51 .62 .48 .64 .73 .71 .86 .75 .83 .89 .88 .74 .57 .69 .79 .67 74.72 .83 .53 Attitude 184 Information retrieval Information storage Information sharing Information application .89 .91 .83 Computers in Human Behavior 90 (2019) 181–187 Construct Computers in Human Behavior 90 (2019) 181–187 I. Arpaci attributes of the students’ information management practices and their attitudes. Behavioral intentions and attitudes have nominal values and classified as low, medium or high by using the Mean ± SD combination. On the other hand, the attributes of information retrieval, storage, sharing, and application have numeric values. The predictive model was tested by using Weka (ver. 3.8.1) based on the most common classification algorithms, which classify data into a number of pre-defined classes with different learning schemas. The results shown in Table 4 provided performance of the best performance classifier based on two test modes; 10-fold cross-validation and percentage split (66.0%). The study employed a linear logistic regression classifier (Logistic), a meta AdaBoostM1, a Bayesian classifier (Naive bayes), a lazy LWL, and two rule learners (OneR, JRip). The results showed that the OneR rule learner performs slightly better than the other classifiers. OneR classifier predicted the behavioral intentions with an accuracy of 73.70% and 71.43% for 10-fold cross-validation and percentage split, respectively. Further, this algorithm had a better performance in terms of true positive (TP) rate (0.74) and F-measure (0.73) for the 10-fold test mode compared to the other classifiers (i.e. Naive bayes, Logistic, JRip, AdaBoostM1, LWL). The results shown in Table 5 suggested that attributes of the attitudes, information retrieval, information storage, information sharing, and information application predicted the behavioral intentions with an accuracy of 74%, and thereby, H1b, H2b, H3b, H4b, and H5b were supported. Table 2 KMO and Bartlett's test results. Behavioral intention Attitude Information retrieval Information storage Information sharing Information application KMO Chi-Square Sig. .88 .88 .83 .75 .65 .71 1022.27 1001.80 839.40 594.58 198.79 357.46 .001 .001 .001 .001 .001 .001 (1951) “test of sphericity” suggested that conducting a CFA on the data was appropriate (See Table 2). KMO values were all above 0.65 and Bartlett's test values were all significant (p < .001). This suggested that measures for each construct were interdependent (Leech, Barrett, & Morgan, 2005). The CFA was conducted using maximum-likelihood estimates in an attempt to test the research model. Results suggested the structural model demonstrates a “good fit” to data: [χ2/df = 1.67, GFI = 0.91, NFI = 0.92, AGFI = 0.88, CFI = 0.96, TLI = 0.96, IFI = 0.96, RMSEA = 0.05]. A GFI, NFI, TLI, CFI, and IFI equal or above 0.90 and a RMSEA less than 0.08 indicate an acceptable model fit (Hair et al., 2006). Further, Kline (2005) suggested a χ2/df less than 2.0 indicates a good fit. 5.3. Hypothesis testing 6. Discussion and conclusion A SEM approach was used to explore causal relationships among the variables. The standardized-path-coefficients in the structural model were all statistically significant (See Fig. 1). Results indicated that information retrieval, information storage, information sharing, and information application were significantly associated with the attitudes toward using the MCC services. All these factors explained .33 of the variance in the attitudes. Whereas, the information retrieval had the largest impact on the attitudes with a standardized coefficient of 0.45. On the other hand, the information storage had a relatively small impact on the attitudes with a standardized coefficient of 0.16. The results shown in Table 3 suggested the attitudes were significantly associated with the behavioral intentions with a standardized coefficient of 0.77. These factors accounted for a significant variance in explaining the behavioral intentions (R2 = 0.59). 6.1. Research implications Using the MCC services provides students with ubiquitous access to instructional materials and information anytime and anywhere. In addition, these services provide students with a storage capacity to store their reference materials. Furthermore, the MCC services enable the students to share ideas, documents, and files with peers and classmates. Eventually, the students can apply and transform that knowledge and experience in problem-solving and decision-making. However, the decision to integrate these services in educational settings requires determining an optimal use scenario. This study therefore focused on the use of the MCC services for personal information management. Accordingly, the study employed a hybrid modeling approach for investigating the role of students’ information management (i.e. retrieve, store, share, and apply) practices in their attitudes (toward use) and the intentions to (continue to) use the MCC services. The actuality of the ideas and employing an integrated multi-analytical approach were the novel contributions of the study. The study developed a research model by extending the TRA with additional constructs and tested the model by integrating the SEM and machine learning approach. This study identified the causal relationships between the dependent and independent variables by employing the SEM approach. The SEM results indicated that the students’ information management practices (i.e. retrieve, store, share, and apply) and attitudes were significantly associated with the behavioral intentions to use the MCC services for educational purposes. In a complementary study, machine learning algorithms were 5.4. Testing the predictive model using the machine learning algorithms Classification is one of the machine learning techniques that aims to build a classifier model in order to predict a behavior through classifying the data into a number of pre-defined classes based on a certain criterion (Ngai, Xiu, & Chau, 2009). Common techniques employed for the classification are Bayesian networks, decision trees, if-then-else rules, and neural networks. Machine learning, an artificial/computational intelligence technique, uses a set of attributes and searches for correlations between the attributes and performance of the learning algorithms (Kotsiantis, Zaharakis, & Pintelas, 2007). This study employed machine learning classification algorithms to predict the behavioral intentions based on Table 3 The SEM results. Hypothesis Path Estimate Std. Estimate Std. Error Critical Ratio p-value Remarks H1a H2a H3a H4a H5a A→BI IR→A ISt→A ISh→A IA→A .983 .412 .084 .125 .207 .776 .446 .156 .180 .279 .098 .059 .030 .043 .045 10.04 6.976 2.802 2.918 4.605 .001 .001 .005 .004 .012 Supported Supported Supported Supported Supported Note: A = Attitude; BI= Behavioral Intention; IR= Information Retrieval; ISt = Information Storage; ISh = Information Sharing; IA= Information Application. 185 Computers in Human Behavior 90 (2019) 181–187 I. Arpaci Table 4 Detailed results for the OneR rule learner. 10-fold cross-validation Avg. Percentage split (66%) Avg. TP Rate FP Rate Precision Recall F-Measure MCC ROC Area PRC Area Class .648 .845 .395 .737 .650 .800 .400 .714 .089 .430 .042 .297 .118 .429 .056 .316 .687 .770 .607 .728 .565 .789 .545 .711 .648 .845 .395 .737 .650 .800 .400 .714 .667 .806 .479 .728 .605 .794 .462 .711 .571 .434 .427 .465 .505 .374 .394 .402 .780 .708 .677 .720 .766 .686 .672 .699 .526 .748 .324 .638 .434 .764 .304 .636 Low Medium High 6.2. Practical implications Table 5 Performance results for the applied classifiers. Algorithm Naive Bayes Logistic JRip AdaBoostM1 LWL a b c CCIa (%) TPb-rate Low Medium High Precision An effective management of information has a potential to improve the students’ academic performance and efficiency of the learning tasks. This implies that employing the MCC services for personal information management should be supported and encouraged in the higher education by designing authentic learning environments and scaffolding students in using such services. Universities should develop their own policies (i.e. bring-your-own-device policy) and cloud-based applications to maximize benefits of the such services. More importantly, instructors should employ blended learning strategies that require information management practices to promote educational use of the MCC services. F-Measure 10-fold PSc 10-fold PS 10-fold PS 10-fold PS 67.86 72.08 72.40 71.43 72.08 64.76 71.43 71.43 70.48 70.48 .679 .724 .714 .705 .721 .648 .714 .781 .781 .705 .664 .712 .711 .604 .706 .647 .720 .763 .763 .683 .666 .713 .711 .650 .689 .642 .696 .771 .771 .677 CCI: Correctly Classified Instances. TP: True Positive. PS: Percentage Split. 6.3. Limitations and future directions employed to explore whether the students' information management practices and attitudes would predict the behavioral intentions. The study tested a number of classification algorithms with different methodologies, including bayesian networks, decision trees, if-then-else rules, association rules, and neural networks. The results indicated that the OneR rule learner performs slightly better than other classifiers. It is worthy to mention that OneR (or one class) algorithm works well if one of the attributes can predict the outcome (dependent) variable better than the other attributes in the training data (Holte, 1993). The classifier model suggested that the attitude predicts the behavioral intention better than the other attributes (CCI = 0.74). In line with these results, the structural model suggested that the attitude was a significant factor in predicting the behavioral intention (β = 0.77). Further, the SEM results suggested the students’ information management practices and attitudes were significantly associated with the behavioral intentions and accounted for 59% of the variance in the behavioral intentions. It is important to note that the relation between the attitudes and the behavioral intentions explains a lot of variance in the model, which is desirable, but not informative or useful. Conceptually, the attitudes (toward use) and the intentions to (continue to) use are a (conceptual) component of the later. However, this study overcomes this design flaw by capitalizing on its two main strengths: 1) the complementary use of parametric and nonparametric approaches, and 2) the extension of the original theory (TRA) into the MCC area. The SEM model included information management practices as exogenous (independent) variables. In the model, both the attitudes and behavioral intentions were endogenous (dependent) variables. The SEM results identified the causal relationships among the endogenous variables and between the endogenous and exogenous variables. On the other hand, the classifier model included the attitudes along with information management practices as predictors of the behavioral intentions. Although, presentation of the two analyses (SEM vs. machine learning) seems two separate results, considering the operationalization of the same constructs in both model, indeed, this study presents a “hybrid modeling” approach and not a mere dual-perspective approach. The SEM does not provide a complete solution to all statistical problems of the prediction. On the other hand, it provides a powerful means of testing a causal model based on a well-specified conceptual theory. Accordingly, the complementary use of the SEM and the machine learning algorithms to develop a prediction model provided unique information on the performance of the SEM-based prediction compared to the other prediction models. Besides, using two approaches, one focused on contrasting/comparing and another on completing, help make robust predictions. However, the study has a number of limitations. This study did not account for the potential moderating role of cultural orientations in the educational technology acceptance, as reported by (Arpaci, 2015; Arpaci, Cetin Yardimci, & Turetken, 2015). Therefore, one should be careful to generalize the findings to the samples with a different sociocultural background. Concerns reported by (Stergiou & Psannis, 2017) may also affect the degree of adoption. Future studies may therefore focus on the concerns of confidentiality, privacy, latency, performance instability, data security, and lack of service-level-agreements. The fact that an effective technology integration into the higher education requires a techno-pedagogical approach and needs to consider the aspects of pedagogy, curriculum, change management, competencies, and organizational readiness, implies that future studies should take into account all these aspects. Finally, a longitudinal study is recommended to overcome limitations of the cross-sectional design. References Aepona (2018). Mobile cloud solutions. Retrieved from http://www.aepona.com/ solutions/mobile-cloud/, Accessed date: 16 July 2018. Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50(2), 179–211. Alavi, M., & Leidner, D. E. (2001). Review: Knowledge management and knowledge management systems: Conceptual foundations and research issues. MIS Quarterly, 25(1), 107–136. Arpaci, I. (2015). A comparative study of the effects of cultural differences on the adoption of mobile learning. British Journal of Educational Technology, 46(4), 699–712. https://doi.org/10.1111/bjet.12160. Arpaci, I. (2016). Understanding and predicting students' intention to use mobile cloud storage services. Computers in Human Behavior, 58, 150–157. https://doi.org/10. 1016/j.chb.2015.12.067. 186 Computers in Human Behavior 90 (2019) 181–187 I. Arpaci (pp. 66–78). Hershey: IRM Press. Ngai, E. W., Xiu, L., & Chau, D. C. (2009). Application of data mining techniques in customer relationship management: A literature review and classification. Expert Systems with Applications, 36(2), 2592–2602. Nistor, N. (2014). When technology acceptance models won't work: Non-significant intention-behavior effects. Computers in Human Behavior, 34, 299–300. https://doi.org/ 10.1016/j.chb.2014.02.052. Nonaka, I. (1994). A dynamic theory of organizational knowledge creation. Organization Science, 5(1), 14–37. Nonaka, I., & Takeuchi, H. (1995). The knowledge-creating company: How Japanese companies create the dynamics of innovation. New York: Oxford University Press. Ogiela, M. R. (2017). Impact of ubiquitous computing technologies for modern information management. International Journal of Information Management, 1(37), 1517–1518. Park, E., & Kim, K. J. (2014). An integrated adoption model of mobile cloud services: Exploration of key determinants and extension of technology acceptance model. Telematics and Informatics, 31(3), 376–385. Rogers, E. M. (2003). Diffusion of innovations (5th ed.). New York, NY: Free Press. Sharma, S. K., Al-Badi, A. H., Govindaluri, S. M., & Al-Kharusi, M. H. (2016a). Predicting motivators of cloud computing adoption: A developing country perspective. Computers in Human Behavior, 62, 61–69. Sharma, S. K., Joshi, A., & Sharma, H. (2016b). A multi-analytical approach to predict the Facebook usage in higher education. Computers in Human Behavior, 55, 340–353. Shiau, W. L., & Chau, P. Y. (2016). Understanding behavioral intention to use a cloud computing classroom: A multiple model comparison approach. Information & Management, 53(3), 355–365. Shon, T., Cho, J., Han, K., & Choi, H. (2014). Toward advanced mobile cloud computing for the internet of things: Current issues and future direction. Mobile Networks and Applications, 19(3), 404–413. Stergiou, C., & Psannis, K. E. (2017). Recent advances delivered by mobile cloud computing and internet of things for big data applications: A survey. International Journal of Network Management, 27(3) e1930. Subashini, S., & Kavitha, V. (2011). A survey on security issues in service delivery models of cloud computing. Journal of Network and Computer Applications, 34(1), 1–11. Tan, G. W. H., Ooi, K. B., Leong, L. Y., & Lin, B. (2014). Predicting the drivers of behavioral intention to use mobile learning: A hybrid SEM-neural networks approach. Computers in Human Behavior, 36, 198–213. Turban, E., Sharda, R., & Delen, D. (2011). Decision support and business intelligence systems. Boston: Prentice Hall. Ul Amin, R., Inayat, I., Shahzad, B., Saleem, K., & Aijun, L. (2017). An empirical study on acceptance of secure healthcare service in Malaysia, Pakistan, and Saudi Arabia: A mobile cloud computing perspective. Annals of Telecommunications, 72(5–6), 253–264. Venkatesh, V., Morris, M. G., Davis, G. B., & Davis, F. D. (2003). User acceptance of information technology: Toward a unified view. MIS Quarterly, 27(3), 425–478. Wang, C. S., & Huang, Y. M. (2016). Acceptance of cloud services in face-to-face computer-supported collaborative learning: A comparison between single-user mode and multi-user mode. Innovations in Education & Teaching International, 53(6), 637–648. Arpaci, I. (2017). Antecedents and consequences of cloud computing adoption in education to achieve knowledge management. Computers in Human Behavior, 70(5), 382–390. https://doi.org/10.1016/j.chb.2017.01.024. Arpaci, I., Cetin Yardimci, Y., & Turetken, O. (2015). A cross-cultural analysis of smartphone adoption by Canadian and Turkish organizations. Journal of Global Information Technology Management, 18(3), 214–238. https://doi.org/10.1080/1097198X.2015. 1080052. Bartlett, M. S. (1951). The effect of standardization on a χ2 approximation in factor analysis. Biometrika, 38(3/4), 337–344. Benbasat, I., & Barki, H. (2007). Quo vadis TAM? Journal of the Association for Information Systems, 8(4), 211–218. Chen, S. L., Chen, J. H., & Lee, Y. H. (2018). A comparison of competing models for understanding industrial organization's acceptance of cloud services. Sustainability, 10(3), 673. Cuervo, E., Balasubramanian, A., Cho, D. K., Wolman, A., Saroiu, S., Chandra, R., et al. (2010). MAUI: Making smartphones last longer with code offload. Proceedings of the 8th international conference on Mobile systems, applications, and services (pp. 49–62). ACM. Davenport, T. H., & Prusak, L. (2000). Working knowledge: How organizations manage what they know. Boston, MA: Harvard Business School Press. Davis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly, 13(3), 319–340. Dinh, H. T., Lee, C., Niyato, D., & Wang, P. (2013). A survey of mobile cloud computing: Architecture, applications, and approaches. Wireless Communications and Mobile Computing, 13(18), 1587–1611. Field, A. (2005). Discovering statistics using SPSS. Thousand Oaks, CA: SAGE. Fishbein, M., & Ajzen, I. (1975). Belief, attitude, intention and behavior: An introduction to theory and research. Reading, MA: Addison-Wesley. Hair, J. F., Tatham, R. L., Anderson, R. E., & Black, W. (2006). Multivariate data analysis (6th ed.). Upper Saddle River, NJ: Pearson Prentice Hall. Holte, R. C. (1993). Very simple classification rules perform well on most commonly used datasets. Machine Learning, 11(1), 63–90. Kaiser, H. F. (1970). A second generation little jiffy. Psychometrika, 35(4), 401–415. Kim, S. H., & Kim, J. K. (2018). Determinants of the adoption of mobile cloud computing services: A principal-agent perspective. Information Development, 34(1), 44–63. Kline, R. B. (2005). Principles and practice of structural equation modeling (2nd ed.). New York: The Guilford Press. Kotsiantis, S. B., Zaharakis, I., & Pintelas, P. (2007). Supervised machine learning: A review of classification techniques. Emerging Artificial Intelligence Applications in Computer Engineering, 160, 3–24. Lee, Y. C. (2016). Why do people adopt cloud services? Gender differences. Social Science Information, 55(1), 78–93. Leech, N. L., Barrett, K. C., & Morgan, G. A. (2005). SPSS for intermediate statistics: Use and interpretation (2nd ed.). Hillsdale, NJ: Lawrence Erlbaum. Lian, J. W., Yen, D. C., & Wang, Y. T. (2014). An exploratory study to understand the critical factors affecting the decision to adopt cloud computing in Taiwan hospital. International Journal of Information Management, 34(1), 28–36. Mitchell, H. J. (2003). Technology and knowledge management: Is technology just an enabler or does it also add value? Knowledge management: Current issues and challenges 187