Infrastructure Elements for Smart Campuses: A Bibliometric Analysis

: Sustainable development can be attained at a microlevel and having smart campuses around the world presents an opportunity to achieve city-wide smartness. In the process of attaining smartness on campuses, the elements requiring attention must be investigated. There are many publications on smart campuses, and this investigation used the bibliometric analysis method to identify such publications produced over the last decade. A matrix of 578 nodes and 3217 edges was developed from 285 publications on smart campus construction and procurement. Fifteen cluster themes were produced from the bibliometric analysis. The ﬁndings revealed that China contributed 48.4% of all published articles on the smart campus. The ﬁndings presented a framework from the cluster themes under the four broad infrastructure areas of building construction or repurposing, technology and IT network, continuous improvement, and smart learning and teaching management. The implications of the ﬁndings identiﬁed that IT project management, traditional procurement strategy, and standard forms of contracts such as the New Engineering Contract (NEC) and the Joint Contract Tribunal (JCT) are applicable in the procurement of smart cities.


Introduction
The emergence of smart campuses worldwide has become an opportunity for educational institutions to enhance their utilisation of existing physical infrastructure to introduce smart technologies [1]. The construction and upgrading of smart campuses have seen the deployment of various infrastructures such as sensors, microgrids, smart classrooms; smart parking; smart meters; energy consumption management and digital technologies to improve learning and teaching on campuses [1,2]. Moreover, smart campuses are essential for meeting the United Nations' sustainable development goals (SDGs) like affordable and clean energy and sustainable cities and communities [1]. Likewise, campuses are smaller communities within cities that may serve as micro smart cities and an available framework for smart city development [3,4]. Awuzie et al. [3] suggested that smart cities development in developing countries can be hinged on a clear strategic framework inclusive of governance, policymaking, and management of campuses.
Similarly, Verstaevel et al. [5] suggested that 70% of the world's population will be living in urban areas by the year 2050, and the development of smart campus information technology (IT) applications is an integral part of creating smart campuses. The need for smart cities worldwide is predicated on the problems of the epidemic, climate change, unhealthy cities and access to modern amenities such as the internet of things [1,3]. The development of smart cities is viewed as expensive, non-inclusive of low-income earners, but Awuzie et al. [3] argued that the microlevel application of smart infrastructure through Table 1. Publication extraction protocol.

Source
Input Variables
The second stage of this analysis produced 538 outputs in conference proceedings, journal articles, survey data, books, book chapters on smart campus technologies, and construction. An estimated 538 articles were profiled for suitability. Upon reviewing the outputs, 285 relevant articles were selected and transferred to produce the matrix for the data analysis. The reduction of the articles from Web of Science and Scopus used the input variables in Table 1 to evaluate the content of the abstracts to filter out publications outside the scope of the input variables. Duplicate publications were also eliminated from the list of publications.
In the third stage of the analysis, the matrix produced nodes and edges for application in the Gephi software. In all, 578 nodes and 3217 edges were developed in the matrix. The nodes are the keywords in a network, and the edges represent connecting lines for each node within a network [25][26][27]. The nodes and edges were derived from the abstracts and keywords.
The fourth step made use of the nodes and edges tables in the matrix file to visualise the smart campus technologies in Gephi. This process led to identifying the trends in smart campus research, the geographical distribution of the publications and emerging themes from the analysis. The themes led to developing the framework containing the infrastructure elements for a smart campus, which will be essential for procurement and construction purposes.

Data Analysis and Visualisation
The dataset was visualised using Gephi software, and the analyses covered the strengths of keywords of smart campus technologies, construction and procurement [25][26][27]. The modularity of the nodes and harmonic closeness centrality were used to categorise similar and dominant nodes, thereby creating the cluster themes. Before this process, the overview of trends in smart campus research and the geographical distribution of the publications were presented in Section 3 to understand the growth and location of smart campus studies worldwide.

Trend in Smart Campus Research
Out of the 285 documents reviewed and analysed in this study in the last ten years, less than ten articles were published on smart campus technologies and construction between 2011 and 2014, as illustrated in Figure 1.

Data Analysis and Visualisation
The dataset was visualised using Gephi software, and the analyses covered the strengths of keywords of smart campus technologies, construction and procurement [25][26][27]. The modularity of the nodes and harmonic closeness centrality were used to categorise similar and dominant nodes, thereby creating the cluster themes. Before this process, the overview of trends in smart campus research and the geographical distribution of the publications were presented in Section 3 to understand the growth and location of smart campus studies worldwide.

Trend in Smart Campus Research
Out of the 285 documents reviewed and analysed in this study in the last ten years, less than ten articles were published on smart campus technologies and construction between 2011 and 2014, as illustrated in Figure 1.  The trends in smart campus publications from 2017 indicate more awareness and desire to study smart campus applications. Publications from 2021 are still in the growth process, with 25 articles far greater than the combination of smart-campus-related articles from 2011 to 2014. The non-inclusion of these 25 articles in this analysis will prove a major deficit in discussing new technologies related to smart campuses. This result denotes a greater need for smart campus development. With the growth in smart campus research, the geographical distribution of the articles, as illustrated in Figure 2, indicated that 138 articles emanated from China. They are forming 48.4% of the total sample size. The key articles from China are Wang [23]; Bastidas-Manzano [24]; Rochat [25]; and Thangaraj [26] that featured localised and regional studies on the construction of smart campuses using big data architecture, cloud computing, learning analytics and teaching systems. The trends in smart campus publications from 2017 indicate more awareness and desire to study smart campus applications. Publications from 2021 are still in the growth process, with 25 articles far greater than the combination of smart-campus-related articles from 2011 to 2014. The non-inclusion of these 25 articles in this analysis will prove a major deficit in discussing new technologies related to smart campuses. This result denotes a greater need for smart campus development. With the growth in smart campus research, the geographical distribution of the articles, as illustrated in Figure 2, indicated that 138 articles emanated from China. They are forming 48.4% of the total sample size. The key articles from China are Wang [23]; Bastidas-Manzano [24]; Rochat [25]; and Thangaraj [26] that featured localised and regional studies on the construction of smart campuses using big data architecture, cloud computing, learning analytics and teaching systems. Other dominant studies emerged from the United States of America, Taiwan, Italy, Brazil, and Canada with 34, 27, 23, 18 and 14 publications respectively. The combination of these countries only accounts for a combined percentage of 89.12% of the total publications of smart campus research, while other countries account for only 10.88% publication in this field of research. The focus of the publication search was to identify smart campus technologies and procurement indicators. This preliminary search analysis provided a basis for categorising smart campus infrastructure for procurement purposes.

Categorising Smart Campus Infrastructure
Gephi is a social network analysis tool used to visualise the strength of connections between nodes [29]. The 578 nodes and 3217 edges produced the modularity class used to filter and rank nodes with a resolution of 0.480 accurately. Modularity class measures the strength of the network structure for commonalities and clusters [29,30]. The modularity analysis in this study produced 15 commonalities, as indicated in Figure 3. Correspondingly, the initial social network map as given in Figure 4 shows the combination of all 15 modularity classes. Other dominant studies emerged from the United States of America, Taiwan, Italy, Brazil, and Canada with 34, 27, 23, 18 and 14 publications respectively. The combination of these countries only accounts for a combined percentage of 89.12% of the total publications of smart campus research, while other countries account for only 10.88% publication in this field of research. The focus of the publication search was to identify smart campus technologies and procurement indicators. This preliminary search analysis provided a basis for categorising smart campus infrastructure for procurement purposes.

Categorising Smart Campus Infrastructure
Gephi is a social network analysis tool used to visualise the strength of connections between nodes [29]. The 578 nodes and 3217 edges produced the modularity class used to filter and rank nodes with a resolution of 0.480 accurately. Modularity class measures the strength of the network structure for commonalities and clusters [29,30]. The modularity analysis in this study produced 15 commonalities, as indicated in Figure 3. Correspondingly, the initial social network map as given in Figure 4 shows the combination of all 15 modularity classes.
The modularity class from Figure 3 was extracted and presented in Table 2 in terms of individual class percentage distribution in the network.
Modularity class 5 contains 21.80% of the network and has the cluster theme as being a smart campus network grid. Internet of things (IoTs) and smart buildings cluster theme is drawn from a 13.15% modularity class. 7.44% of the modularity produced the cloud computing cluster theme. Likewise, the campus information portal theme has a 7.44% modularity. Deep learning architecture and campus Equipment Management Services (CEMS) each have a percentage of 5.71%. Data mining is another major theme containing 5.54% of the modularity. Smart city, inclusive smart technology, and applications cluster themes each contribute 4.67% to the network. 3.29% of the cluster theme is the autoanalysers, and 3.29% of the network contains performance measurement and forecasting. The energy management system and education management system have 2.77% and 2.25% individually. Correspondingly, the cluster year of the publications in this analysis is between 2012 and 2020. The commonalities between the individual modularities from the social network map in Figure 4 used the most relevant harmonic closeness centrality of the edges connecting the nodes in each module.  The modularity class from Figure 3 was extracted and presented in Table 2 in terms of individual class percentage distribution in the network.   The modularity class from Figure 3 was extracted and presented in Table 2 in terms of individual class percentage distribution in the network.   The harmonic closeness centrality measures are a social network analysis modification of closeness centrality, which uses index values of the most central larger nodes [31,32]. Thus, the closeness of the nodes was measured, and indices were derived. Tables 3-13 present the harmonic closeness centrality measures for each node within a modularity network, the individual nodes forming the clusters and source of publications with the year. Sections 4.1 and 4.2 discusses each smart campus cluster theme from Table 2 in order of their modularity numbers to support the framework in Section 5.

Auto-Analysers
The term "auto-analysers" was coined out of the nodes in modular class 0, which consists of air quality control monitoring with harmonic closeness centrality measure of 0.458, in campus buildings; the application of machine learning (0.422) and experiments (0.422) to understand the dynamic performance of buildings [1,21]. Six key journal articles and two conference papers produced the drivers for the auto-analyser theme and the average year of publication is 2018. Hence, auto-analysers are still new in smart campus research as indicated in Table 3. In terms of auto-analysers, the inclusion of unsupervised machine learning architecture to evaluate the performance of smart quality is an important component of smart buildings on campuses.
Machine learning, LoRaWAN and Air quality monitoring are the key drivers of the auto analysers group by appearing in 27.3%, 22.7% and 18.2% of relevant documents. Min-Allah and Alrashed [1] provided an example of a smart campus sketch where sensors designed by Google's sidewalk Labs were deployed to monitor air quality in buildings on Campus in Saint Lois University, in the United States. As indicated in Table 3 and Figure 5, the packet error rate used to test the performance of an access receiver can be used along with testbeds to assess the performance of campus buildings [33,34]. The application of LoRaWaN, a point-to-multipoint networking protocol for encryption and identification, and wireless connecting battery-operated devices as a low power wide area network provided an opportunity to extract and assess data from air-quality monitoring devices [35]. Having a smart grid on campus with auto-analysers provides a foundation for smart campus development.   [1] provided an example of a smart campus sketch where sensors designed by Google's sidewalk Labs were deployed to monitor air quality in buildings on Campus in Saint Lois University, in the United States. As indicated in Table 3 and Figure  5, the packet error rate used to test the performance of an access receiver can be used along with testbeds to assess the performance of campus buildings [33,34]. The application of LoRaWaN, a point-to-multipoint networking protocol for encryption and identification, and wireless connecting battery-operated devices as a low power wide area network provided an opportunity to extract and assess data from air-quality monitoring devices [35]. Having a smart grid on campus with auto-analysers provides a foundation for smart campus development.

Energy Management System (EMS)
The development of auto-analysers as part of a smart campus network grid can be integrated with an energy management system (EMS). The EMS cluster theme was pro-

Energy Management System (EMS)
The development of auto-analysers as part of a smart campus network grid can be integrated with an energy management system (EMS). The EMS cluster theme was produced from modularity class 1, as shown in Figure 6 and Table 4. In Table 4, the dominant drivers are renewable energy with harmonic closeness centrality measure of 0.396; BIM to BEM (0.399); Photovoltaic (0.399); energy renovation (0.397); laser scanner (0.394); and net-zero energy building (0.359). All the publications that produced this theme were journals except for one conference article and the cluster year of publication is 2015.  Renewable energy and Photovoltaic devices as the main drivers of the energy management system continually recurred in 25% of analysed documents for this group. Netzero energy buildings consume energy created on-site, mostly through renewable sources [56,57]. Net-zero energy buildings or retrofitted existing buildings on campuses can use photovoltaic solar panels and renovate an existing building to become energy efficient [58]. In achieving an EMS system on campus, building information modelling (BIM) can be applied in building energy model (BEM) when retrofitting or renovating existing campus buildings [59,60]. EMS can work effectively in a smart campus grid where there are machine learning-enabled analysers and cloud computing.

Cloud Computing
Cloud computing is a major component of any smart campus [9,10,39]. Cloud computing as indicated in Figure 7 is an unmanned storage database that is readily available to multiple users [40,41]. The creation of this cluster theme made use of important drivers such as biometric security with harmonic closeness centrality of 0.420; smart university, 0.442; adaptive learning, 0.478; crowd detection, 0.478; architecture, 0.474; Participatory sensing, 0.401; Local binary pattern, 0.396; Sustainable development goals (SDGs), 0.360; Smart community; and smart features, 0.358. The theme also came from a cluster of conference papers published mostly in the year 2020. Renewable energy and Photovoltaic devices as the main drivers of the energy management system continually recurred in 25% of analysed documents for this group. Net-zero energy buildings consume energy created on-site, mostly through renewable sources [56,57]. Net-zero energy buildings or retrofitted existing buildings on campuses can use photovoltaic solar panels and renovate an existing building to become energy efficient [58]. In achieving an EMS system on campus, building information modelling (BIM) can be applied in building energy model (BEM) when retrofitting or renovating existing campus buildings [59,60]. EMS can work effectively in a smart campus grid where there are machine learning-enabled analysers and cloud computing.

Cloud Computing
Cloud computing is a major component of any smart campus [9,10,39]. Cloud computing as indicated in Figure 7 is an unmanned storage database that is readily available to multiple users [40,41]. The creation of this cluster theme made use of important drivers such as biometric security with harmonic closeness centrality of 0.420; smart university, 0.442; adaptive learning, 0.478; crowd detection, 0.478; architecture, 0.474; Participatory sensing, 0.401; Local binary pattern, 0.396; Sustainable development goals (SDGs), 0.360; Smart community; and smart features, 0.358. The theme also came from a cluster of conference papers published mostly in the year 2020.
Cloud computing on campuses provides a strong foundation for smart cities. The smart university is the most predominant indicator of cloud computing through appearing on more than 32% of relevant documents to cloud computing, as shown in Table 5. Cloud computing is integral to achieving SDGs where cleaner and smart campus communities can be developed as prototypes of smart cities. Cloud computing also depends on the architecture and construction of smart university buildings, and can be useful in storing data for crowd detection and enhancing the smartness of campus buildings. Sustainability 2021, 13, x FOR PEER REVIEW 10 of 34 Cloud computing on campuses provides a strong foundation for smart cities. The smart university is the most predominant indicator of cloud computing through appearing on more than 32% of relevant documents to cloud computing, as shown in Table 5. Cloud computing is integral to achieving SDGs where cleaner and smart campus communities can be developed as prototypes of smart cities. Cloud computing also depends on the architecture and construction of smart university buildings, and can be useful in storing data for crowd detection and enhancing the smartness of campus buildings. The existence of a smart city makes it easier for the development of a smart campus. Figure 8 and Table 6 combined the drivers or urban planning with the harmonic closeness centrality indices of 0.394; Arduino, 0.400; global positioning system (GPS), 0.414; data

Smart City
The existence of a smart city makes it easier for the development of a smart campus. Figure 8 and Table 6     As presented in Table 6, Data fusion (25%), Smart mobility (18.8%) and GPS (18.8) are the significant drivers of the Smart city cluster, as shown in Table 6. The main feature of smart cities is the tracking and reporting system available to monitor vehicular movement and the urban environment. SIM808 combined the features of GPS with and complete quad-band global system for mobile communications to track the location of vehicles and people instantaneously [73][74][75][76]. SIM808 is a smart grid along with Arduino capable of creating interactive electronic objects such as text, audio, animation and other forms of visualisation that are important to smart city development [74][75][76]. A smart campus can exist seamlessly within a smart city if the local authority provides the required elements and infrastructure.

Education Management System
The education management system was produced from the nodes in Figure 9 (drivers in The education management system was produced from the nodes in Figure 9 (drivers in Table 7) from mostly conference papers published in 2012. The drivers with harmonic closeness centrality indices are education management, 0.399; E-learning, 0.473; artificial intelligence, 0.412; education 3.0 framework, 0.396; mobile computing, 0.393; mobile learning, 0.395; and key performance indicators, 0.390.  An education management system in a smart campus is an important component of smart infrastructure procurement. E-learning, Mobile computing and Artificial intelligence are the critical indicators of the education management system by being recapped on 30.8%, 23.1%, and 9.2% of related documents to the education management system. A smart campus requires mobile learning and computing adoption because students and academics already have smart mobile devices [15,[81][82][83]. Therefore, the need to create an educational framework that identifies key performance indicators of smart and electronic learning hinged on artificial intelligence and mobile computing. An education management system in a smart campus is an important component of smart infrastructure procurement. E-learning, Mobile computing and Artificial intelligence are the critical indicators of the education management system by being recapped on 30.8%, 23.1%, and 9.2% of related documents to the education management system. A smart campus requires mobile learning and computing adoption because students and academics already have smart mobile devices [15,[81][82][83]. Therefore, the need to create an educational framework that identifies key performance indicators of smart and electronic learning hinged on artificial intelligence and mobile computing.

Smart Campus Network Grid
From the initial network analysis in Figure 4, the smart campus is the largest node. Relatedly, Figure 10 extracted the network analysis for modular class 5. The smart campus network grid in Table 8 depends on the smart campus construction, which is supported by the harmonic closeness centrality indices of 0.683 and 0.398 for smart campus and construction correspondingly. The cluster contributing articles are mainly from conference papers published within 2018. From the initial network analysis in Figure 4, the smart campus is the largest node. Relatedly, Figure 10 extracted the network analysis for modular class 5. The smart campus network grid in Table 8 depends on the smart campus construction, which is supported by the harmonic closeness centrality indices of 0.683 and 0.398 for smart campus and construction correspondingly. The cluster contributing articles are mainly from conference papers published within 2018.   As shown in Table 8, Smart campus appeared on 95.3% of the relevant articles and papers to the Smart campus network grid, which revealed the critical role of smart campus as the prime indicator of this group among other nine indicators. The smart campus network grid is the bedrock of creating a smart campus. Learning institutions need to develop smart buildings on campus [5,57,88,90,93,94]. Betancur et al. [95] suggested social participation in smart cities development. In this instance, end users are considered essential in drawing out plans for smart communities. Students on smart campuses constitute most of the end-user population on campus. Students use smartphones, and it will be easier for learning institutions to adopt the approach of a microgrid, microservice, education big data extraction and analysis. The inclusion of courses such as robotics courses and research can also foster the attainment of a smart campus. Internet of things cannot be separated from a smart campus network grid and will be discussed in the subsequent section.     Internet of things (IoTs) and smart buildings interlock with the smart network grid theme. Figure 11 illustrates the network for this theme, and Table 9 provides the important drivers in terms of their harmonic closeness centrality indices. The dominant driver is artificial intelligence, with an index of 0.641. Energy; Fog computing; 5G and 6G flagship have indices of 0.467, 0.452, 0.425, and 0.401, individually. Smart buildings, green computing, sustainability, and smart grid have their indices as 0.362, 0.401, 0.271, and 0.297, respectively. Likewise, the key sources for this theme emanated from conference papers that were published within 2020. Thus, implying that the studies and applications of the IoTs and smart buildings on campuses worldwide are recent developments. arated from a smart campus network grid and will be discussed in the subsequent section.

Internet of Things (IoTs) and Smart Buildings
Internet of things (IoTs) and smart buildings interlock with the smart network grid theme. Figure 11 illustrates the network for this theme, and Table 9 provides the important drivers in terms of their harmonic closeness centrality indices. The dominant driver is artificial intelligence, with an index of 0.641. Energy; Fog computing; 5G and 6G flagship have indices of 0.467, 0.452, 0.425, and 0.401, individually. Smart buildings, green computing, sustainability, and smart grid have their indices as 0.362, 0.401, 0.271, and 0.297, respectively. Likewise, the key sources for this theme emanated from conference papers that were published within 2020. Thus, implying that the studies and applications of the IoTs and smart buildings on campuses worldwide are recent developments.   Sustainability appeared on 18.2% of documents clustered under the internet of things (IoTs) and smart buildings module. Successively Smart grid, Artificial intelligence and Fog computing occurred in ±10% of relevant documents. Other cluster drivers from Table 9 are outdoor applications, water management, repurposing old campus buildings, persuasive computing, lower power wide area network, smart environment, teaching and learning. IoTs and smart buildings can benefit from 5G and the coming 6G flagship network system [97][98][99][100]. A smart campus will also require new smart buildings or repurposed old campus buildings to meet the needs of a smart environment for teaching and learning [29]. Water management on campus is also an essential component of a smart campus. Waterless urinals may be adopted to save millions of litres of water on campuses. A smart network grid will effectively integrate with smart buildings with IoTs.

Inclusive Smart Technology
Inclusive smart technology in modular class 7 encapsulates the nodes in Figure 12. In Table 10, the inclusion of people with disabilities formed the basis of this cluster theme even though the harmonic closeness centrality index is 0.390. A smart campus with smart buildings needs to manage big data, create fragmented learning, and use ubiquitous gamebased learning and infotainment learning for students with learning difficulties. The drivers mentioned above have harmonic closeness centrality indices of 0.517, 0.397, 0.351, 0.395, individually. Equally, information security with an index of 0.393 is required to create an inclusive smart campus thereby assuring the users of their data safety. This cluster theme was developed from key journal sources published from the year 2018. even though the harmonic closeness centrality index is 0.390. A smart campus with smart buildings needs to manage big data, create fragmented learning, and use ubiquitous game-based learning and infotainment learning for students with learning difficulties. The drivers mentioned above have harmonic closeness centrality indices of 0.517, 0.397, 0.351, 0.395, individually. Equally, information security with an index of 0.393 is required to create an inclusive smart campus thereby assuring the users of their data safety. This cluster theme was developed from key journal sources published from the year 2018.  In the Inclusive smart technology cluster, big data is the primary driver as it appeared on more than 50% of relevant journal articles and conference papers. Min-Allah and Alrashed [1] sketched a template for an inclusive smart campus technology where people services, smart utility, resource management, and educations services formed a good approach of providing inclusive services for the end-users on a campus. Inclusive technology from the perspective of learning can make use of game-based and infotainment learning [105,107]. Fragmented learning breaks down the teaching of topics into unique disciples to include a diverse population of students; this can be made possible [105,107]. Inclusive technology on campuses is important for promoting diversity, equality and management of learning and teaching.

Performance Measurement and Forecasting
Modular class 8 produced performance measurement and forecasting cluster theme out of the nodes in Figure 13 as presented in Table 11. Performance measurement and forecasting are associated with auto-scaling, forecasting, loss models; path loss; horizontal scalability; radio-propagation; quality of service; and vertical scalability. The harmonic closeness centrality indices for the drivers are, respectively, 0.462, 0.460, 0.400, 0.402, 0.442, 0.392, 0.434, and 0.402. These drivers were drawn from a combination of journal articles and survey data published mostly in 2020.  In the Inclusive smart technology cluster, big data is the primary driver as it appeared on more than 50% of relevant journal articles and conference papers. Min-Allah and Alrashed [1] sketched a template for an inclusive smart campus technology where people services, smart utility, resource management, and educations services formed a good approach of providing inclusive services for the end-users on a campus. Inclusive technology from the perspective of learning can make use of game-based and infotainment learning [105,107]. Fragmented learning breaks down the teaching of topics into unique disciples to include a diverse population of students; this can be made possible [105,107]. Inclusive technology on campuses is important for promoting diversity, equality and management of learning and teaching.

Performance Measurement and Forecasting
Modular class 8 produced performance measurement and forecasting cluster theme out of the nodes in Figure 13 as presented in Table 11. Performance measurement and forecasting are associated with auto-scaling, forecasting, loss models; path loss; horizontal scalability; radio-propagation; quality of service; and vertical scalability. The harmonic closeness centrality indices for the drivers are, respectively, 0.462, 0.460, 0.400, 0.402, 0.442, 0.392, 0.434, and 0.402. These drivers were drawn from a combination of journal articles and survey data published mostly in 2020.  As shown in Table 11, each performance measurement and forecasting cluster indicator frequently appeared between 8% to 15% of the relevant documents, while Forecasting as the prime indicator appeared on 23.1% of documents. Performance measurement and forecasting are crucial in understanding the impact of the smart grid network, IoTs, inclusive smart technology and every other smart application on teaching and learning [35,110,111]. In improving smart campus technology, the use of path and loss models and the quality of service can provide more clarity on the direction the smart campus infrastructure is going. Forecasting is also essential to assess the energy performance of campus buildings, the effectiveness of campus technologies in delivering quality services and learning experiences to students. Vertical scalability also helps to resize servers by increasing the power with additional features, while horizontal scalability included more resources and hardware to existing network infrastructures [35,58]. The development of servers and IoTs hardware on campuses will depend on forecasting and the previous performance of the system. Performance measurement and forecasting will create more opportunities for advances in smart campus applications.

Smart Campus Applications
Smart campus applications in modular class 9, as illustrated in Figure 14 Higher education and social networking are the key drivers of the Smart campus applications by appearing on 33.3% and 20% of analysed documents in this cluster. Table  12 focused more on higher education, teaching and learning applications of IT. Mobile applications are useful for timetabling, attendance monitoring, communication, and feed back purposes [2,90,115]. More importantly, mobile applications have become useful in teaching and learning using Microsoft applications. Social network applications can be developed specifically for the campus in the furtherance of the learning experience. Post adoption of smart campus applications can be associated with the performance measure ment and forecasting themes. By so doing, learning management systems can be devel oped and enhanced.  Higher education and social networking are the key drivers of the Smart campus applications by appearing on 33.3% and 20% of analysed documents in this cluster. Table 12 focused more on higher education, teaching and learning applications of IT. Mobile applications are useful for timetabling, attendance monitoring, communication, and feedback purposes [2,90,115]. More importantly, mobile applications have become useful in teaching and learning using Microsoft applications. Social network applications can be developed specifically for the campus in the furtherance of the learning experience. Post-adoption of smart campus applications can be associated with the performance measurement and forecasting themes. By so doing, learning management systems can be developed and enhanced.

Learning Management System
The "learning management system" cluster theme in modularity class 10 stemmed from a combination of conference and journal articles published mainly in 2018, as pre-sented in Figure 15 and Table 13. Learning management systems can be integrated with IoTs and smart building, and smart campus applications. The components of learning management systems depend on wireless fidelity (WI-FI), real-time monitoring system; automation; camera; autonomous vehicles; sensors; wireless sensor networks; and Message Queuing Telemetry Transport (MQTT) devices. The above listed drives possess harmonic closeness centrality measures of 0. 305, 0.447, 0.453, 0.451, 0.436, 0.417, 0.392   As shown in Table 13, Sensors and Security appeared on 18.2% of relevant documents, and Wireless sensor network appeared on 13.6% of analysed journal articles and conference papers. Learning management systems features depend on attendance monitoring, the availability of a good WI-FI system coupled with sensors, wireless sensors, and security features to efficiently harness data from end-users and support the delivery of teaching and learning [92,120,121]. The use of autonomous vehicles for student transportation on campuses is another idea that promotes students' learning experience. An efficient WI-FI system is a foundation for IoTs on campuses. Learning is also enhanced when campus end-users have access to free and efficient WI-FI. Security on campus cannot be  Table 13, Sensors and Security appeared on 18.2% of relevant documents, and Wireless sensor network appeared on 13.6% of analysed journal articles and conference papers. Learning management systems features depend on attendance monitoring, the availability of a good WI-FI system coupled with sensors, wireless sensors, and security features to efficiently harness data from end-users and support the delivery of teaching and learning [92,120,121]. The use of autonomous vehicles for student transportation on campuses is another idea that promotes students' learning experience. An efficient WI-FI system is a foundation for IoTs on campuses. Learning is also enhanced when campus end-users have access to free and efficient WI-FI. Security on campus cannot be separated from learning. Campus end users want to feel safe when they learn. This safety pertains to the internet and physical security. Learning management systems can also be associated with data mining features as expressed in modular class 11.

Data Mining
Data mining is a process of extracting information and identifying patterns from large data sets with the aid of machine learning to perform performance measurement and forecasting [85,[124][125][126]. The data mining cluster theme was produced from drivers with harmonic closeness centrality indices such as data analysis, 0.464; bibliometrics, 0.427; web service; 0.392; ranking analytics, 0.396; and statistics, 0.279. Publications mostly from 2018 and survey data formed the sources of the drivers in Table 14 and Figure 16.
Data analysis as the key driver of the Data mining group appeared on 30.4% of documents, while the other four drivers of this group appeared on 17.4% of manuscripts. The Data mining theme interacts with learning management systems when assessing students' performance regarding their attendance in classes, internet usage; campus applications; student grades; library usage; and staff activities on campus and online. Data mining features are statistics based and makes use of machine learning and other analytics [4,103,127]. Data mining features are essential in all facets of smart campus management and improvement. The campus information portal is another theme that data mining will depend on. separated from learning. Campus end users want to feel safe when they learn. This safety pertains to the internet and physical security. Learning management systems can also be associated with data mining features as expressed in modular class 11.

Data Mining
Data mining is a process of extracting information and identifying patterns from large data sets with the aid of machine learning to perform performance measurement and forecasting [85,[124][125][126]. The data mining cluster theme was produced from drivers with harmonic closeness centrality indices such as data analysis, 0.464; bibliometrics, 0.427; web service; 0.392; ranking analytics, 0.396; and statistics, 0.279. Publications mostly from 2018 and survey data formed the sources of the drivers in Table 14 and Figure 16.

Campus Information Portal
Modularity class 12 combined key sources, mainly conference papers and journal articles published recently in the year 2020. The nodes in Figure 17 and Table 15 identified innovative education; smartphone; automated attendance monitoring; blended learning; energy efficiency; energy consumption; education data mining; collaborative learning; learning analytics; sustainable education; smart classroom; and smart boards as the basis of campus information portal. The individual harmonic closeness centrality indices of the abovelisted drivers are 0.504, 0.406, 0.446, 0.438, 0.481, 0.466, 0.441, 0.452, 0.417, 0.441, 0.368, and 0.392, respectively.
Energy efficiency is the predominant driver of the Campus information portal group as it appears on more than 20% of relevant documents. Energy consumption and efficiency, and education data mining are essential in smart campus development and information portal. It is imperative for campus end-users to know the amount of energy consumed and how they can contribute to achieving energy efficiency on campus. Sustainable education can be met through openness on energy-related information and the inclusion of endusers in attaining energy usage goals. Likewise, innovative learning and blended learning approaches such as online and classroom teaching should be visible for students through a campus portal [102,129]. In improving learning on campus, a smart classroom that comprises smart boards with touch screen capabilities will create an enabling environment for 21st-century learners and teachers. Hence, innovative learning depends on meeting the technological needs of students. In this instance, smartphones are essentially an integral part of smart campuses.

Campus Information Portal
Modularity class 12 combined key sources, mainly conference papers and journal articles published recently in the year 2020. The nodes in Figure 17 and Table 15 identified innovative education; smartphone; automated attendance monitoring; blended learning; energy efficiency; energy consumption; education data mining; collaborative learning; learning analytics; sustainable education; smart classroom; and smart boards as the basis of campus information portal.     Deep learning architecture is an advanced form of artificial neural network for problem-solving. A deep learning architecture theme can be built with data mining processes. Table 16 was developed from the network in Figure 18  Energy efficiency is the predominant driver of the Campus information portal gro as it appears on more than 20% of relevant documents. Energy consumption and e ciency, and education data mining are essential in smart campus development and in mation portal. It is imperative for campus end-users to know the amount of energy c sumed and how they can contribute to achieving energy efficiency on campus. Sustaina education can be met through openness on energy-related information and the inclus of end-users in attaining energy usage goals. Likewise, innovative learning and blen learning approaches such as online and classroom teaching should be visible for stude through a campus portal [102,129]. In improving learning on campus, a smart classro that comprises smart boards with touch screen capabilities will create an enabling en ronment for 21st-century learners and teachers. Hence, innovative learning depends meeting the technological needs of students. In this instance, smartphones are essenti an integral part of smart campuses.

Deep Learning Architecture
Deep learning architecture is an advanced form of artificial neural network for pr lem-solving. A deep learning architecture theme can be built with data mining proces Table 16 was developed from the network in Figure 18   Augmented reality, Android and Face recognition are the key drivers of the Deep learning architecture, which occurred on 30%, 25%, and 20% of related manuscripts to Deep learning architecture. Deep learning architecture depends on features such as a sliding window filter used as an incremental database divided into several partitions; accuracy metrics augmented reality, face recognition and android devices [17,134,135]. The application of deep learning architecture is evident in face recognition on campuses. Face recognition is important for campus security and the efficient running of other services such as transportation and attendance monitoring.

Campus Equipment Management Service (CEMS)
The campus equipment management system (CEMS) in the network displayed in Figure 19 was filtered to produce the drivers in Table 17. The drivers and harmonic closeness centrality indices contributing to the cluster theme of CEMS are campus visitor management service, 0.404; hazardous area management service (HAMS), 0.401; application framework, 0.462; radio-frequency identification (RFID), 0.411; smart devices, 0.416; localisation, 0.425; mobile robotics, 0.403; Petri net modelling, 0.397; and optimal deployment, 0.399. The cluster contributing sources came from journal articles published from the year 2012. CEMS is still emerging, and there are several studies about the development of CEMS in smart campus research. ness centrality indices contributing to the cluster theme of CEMS are campus visitor man agement service, 0.404; hazardous area management service (HAMS), 0.401; applicatio framework, 0.462; radio-frequency identification (RFID), 0.411; smart devices, 0.416; loca isation, 0.425; mobile robotics, 0.403; Petri net modelling, 0.397; and optimal deploymen 0.399. The cluster contributing sources came from journal articles published from the yea 2012. CEMS is still emerging, and there are several studies about the development o CEMS in smart campus research.  As shown in Table 17, RFIS is the predominant driver of the Campus Equipment Management Service (CEMS) cluster as it appears on more than 30% of relevant documents. Zhou [138] noted that "campus information station has become an important part of the construction of smart campus". Requirements of such campus visitor management systems and hazardous area management services are becoming increasingly important for physical and online campus security [84,133,136,137]. Interconnectivity with smart devices such as smartphones, smart meters, smart cameras can use RFID to transmit data and conduct performance measurement through Petri net modelling [140]. CEMS must be integrated with the Smart campus network grid, data mining, and performance measurement and forecasting. The framework in Figure 20 and implications of the smart campus technology further provides more insight into the importance and applications of the findings of this bibliometric analysis.

Implications of Findings
This section presents the findings in a conceptual framework. Further discussions are provided under the implications of findings which were derived from Figure 20. The implications for future research are also presented at the end of this section. The implications of findings in this section contains theoretical applications of the findings for policy formation, development, and enhancement of smart campuses.

Framework Conceptualisation
The findings from Section 4 aided in conceptualising a framework that highlights the infrastructure elements for smart campus development. The 15 cluster themes were categorised into four (4) main divisions. The four divisions are smart building construction or repurposing, technology, and IT network; continuous improvement; and smart learning and teaching systems. Smart building construction or repurposing encapsulates smart city clusters, IoTs and smart buildings, and energy management systems. Technology and IT network division comprise of smart campus network grid; cloud computing; inclusive smart technology; and smart campus applications.
Continuous improvement has a similar IT infrastructure for evaluating the performance of smart campus features. The continuous improvement consists of deep learning architecture, data mining, auto-analysers, and performance measurement and forecasting. The final division in the framework contains a campus information portal, a learning management system; a campus equipment management service (CEMS); and an education management system to produce the smart learning and teaching systems infrastructure. Each of the four divisions will be explained in Section 4.2.

Implications of Findings on Procurement of Smart Campus Infrastructure
The implications of the four categories in Figure 20 on procurement of smart campus infrastructure depends on the geographical location; the readiness of education institution where the campus is located; proximity to smart cities; a strategic vision of the education institution; source and availability of finance; technical know-how to construct and deploy smart cities infrastructure [1,36,37]. The relevant articles on smart campus procurement and construction will be discussed under the explanations of the four main divisions from Sections 4.2.1-4.2.4.

Implications on Smart Campus Building Construction or Repurposing
Min-Allah and Alrashed [1] reported the role of smart cities in the development of smart campuses from the viewpoint of people, prosperity, governance and propagation, which should be in line with the community, infrastructure sustainability, administration and replicability on campuses. Zhou [138] also identified the inclusion and repurposing of teaching, scientific and management services infrastructures in constructing a digital campus. The construction of smart campus buildings depends on the available technology and the existence of a smart city. If the campus is situated within a smart city, Min-Allah and Alrashed's [1] approach to creating a smart campus can be adopted along with the findings of this study which is inclusive of energy management systems in smart buildings. Inversely, the chances of repurposing or retrofitting existing campus buildings with the IoTs is another feasible methodology for the provision of smart buildings. What makes a building smart is the energy consumption and management approach, sustainability, IoTs capabilities, smart meters, water management and interconnectivity to harness data.
Construction or repurposing activities to deliver smart buildings will have to consider pertinent stakeholders, procurement method, contract documentation and arrangement, construction method statement, and information management. Standard procurement and contractual approaches such as traditional procurement and New Engineering Contracts (NEC) may suit smart building construction. For the repurposing of buildings, the Joint Contract Tribunal (JCT) for minor works can be applicable. BIM and GIS have been suggested as good approaches to mapping and constructing campuses [14,53]. The application of BIM and GIS in smart building construction will lay a foundation for easier facilities management using the Construction Operations Building Information Exchange (COBie) data.

Implications on Technology and IT Network in a Smart Campus
As suggested in Section 4.2.1 above, the IoTs forms the core of smart buildings. A smart network grid consisting of a microgrid on campus is an integral part of the smart campus building construction [1,132]. The existence of a smart campus network grid will be inclusive of cloud computing. Cloud computing infrastructure provides a single-point solution to solving the challenges of smart campus messaging, data processing, business interfaces, human-computer interaction and persuasive computing [132].
Cloud computing within an IT network must be all-encompassing in supporting the learning and teaching management system whereby students with disabilities are provided with the necessary tools to learn. Likewise, academic staff with disabilities must be included by creating campus-specific applications such as social networks, mobile applications, voice recognition or text-to-speech teaching aids. The construction of new technology on smart campus deal with IT project management. Consequently, there is a need to study the intricacies of contract, construction, and procurement of IT infrastructure on campuses.

Implications on Continuous Improvement of Smart Campus Infrastructure
The existence of smart campus infrastructures such as smart buildings, energy management systems under construction and IT network divisions in Figure 20 will be of no significance without performance measurement and forecasting, data mining, auto-analysers, and deep learning architecture for continuous improvement. The idea of continuous improvement in smart campus management pertains to learning from existing data, practices and understanding how the systems can be improved [127,128].
Data mining supports harnessing analysis with deep learning architecture of neural networks for understanding campus end-user behaviour and requirements [3,17,134,135,[141][142][143]. Continuous improvement of smart campus infrastructure produces opportunities for research into smart campus applications, IT infrastructure, network, management, and applications [3,144,145]. Hence, a framework for improving the existing systems has used vertical and horizontal scalability, loss models, forecasting; statistics; and accuracy metrics. Similarly, evaluations of air quality and energy usage in smart buildings must support the idea of producing net-zero energy buildings. Sustainability and smart campuses meet at the point of net-zero energy provisions in buildings, whereby smart buildings use renewable energy such as photovoltaic panels to meet all the energy demands in a building [105,146]. In achieving this target, continuous assessment and improvement of smart buildings, technology and IT network, and energy management systems will provide a stronger foundation for sustainable buildings on campuses.

Implications on Smart Learning and Teaching Systems
The governance of a smart campus is hinged on services delivered through learning and teaching. Smart learning and teaching systems division of a smart campus infrastructure depends on on-campus information portals with the attachment of automated attendance; education data mining; collaborative learning; learning analytics; and smart classrooms with smart boards [89,94,122]. The learning management system governs how students and other campus end-users interact with the smart technology, smart buildings, and energy usage on campus while learning.
In learning and teaching on campus, the WI-FI system, mobile applications, and communications mostly through emails must be effective. The campus equipment management system also supports learning and teaching when visitors and material waste are managed. Finally, the education management system produces an opportunity for artificial intelligence to be cohesive with E-learning platforms for students and mobile learning applications. IT project management is required to deliver the infrastructure required for smart learning and teaching. In this infrastructure category, IT networks and applications can be developed either through contractors or the existing IT services on campuses. Other tools such as smart classrooms and boards can be procured and installed by the IT service or specialist contractors.

Implication of Findings for Future Research
The important outcomes of the study, as illustrated in Figure 20, form a foundation for further research into the procurement of smart campuses. The areas for future large-scale research are the following: • Application of BIM in delivering smart campuses • IT project management in smart campus procurement • Investigations into the best procurement strategies for smart campuses • Contractual arrangement in smart campus procurement.
The above-listed research areas should be viewed from the geographical and national outlooks because there will be peculiarities of smart campuses regarding campus sizes, the readiness to manage campuses, availability of technology, and financial capabilities.

Conclusions and Limitations of the Study
This study aimed to develop an inclusive framework for smart campus procurement by reviewing all available literature within the last ten years. The findings show a progression in the development of smart campus studies around the world. Smart campus research is still emerging because smart campuses are still under development around the world. The Asian continent, especially China, has been leading the pace in smart campus research. The delivery of smart campuses depends on construction or repurposing activities, technology, and IT network; smart learning and teaching management system for smart campus governance; and continuous improvement of existing smart campus infrastructure. The existence of smart cities makes it easier for smart campuses to emerge around the world and improve. The limitations of this study are evident in the methodological approach, which is more a review of existing literature, the application of keywords, abstracts, and publications in determining smart campus infrastructure. Qualitative and quantitative methods may identify more smart campus infrastructural elements. Nonetheless, the conceptual framework in this study provides a foundation for large scale studies into how sustainable development of the environment can be attained in smart campuses and a new direction for further studies into procurement practices in smart campus development.