Impact of Fog Computing on Indian Smart-Cities: An Empirical Study

The article introduces a two-stage Structural equation modelling- Articial Neural Network (SEM-ANN) model for the Smart city creation through Fog Computing (FC) and Internet of Things (IOT) by identifying the critical success factors in Indian context. The research article introduces a new factor Fog Computing (FC). Internet of Things (IoT) is again sub-divided into three more factors as Internet of People (IoP), Internet of Services (IoS) and Internet of Energy (IoE) as the independent variables. 13 Smart cities and 379 respondents are involved for this study. The data analysis is done through Structural equation modelling (SEM) and articial neural network (ANN) which measures both the linear and non-linear relationships respectively. From Structural Equation Modeling (SEM) output, it is identied that Internet of Things (IOT), Internet of People (IOP) and Internet of Services (IOS) have some signicant positive effect on Fog Computing (FC). Internet of Energy (IOE) has the negative effect on Fog Computing (FC) which is the only exception in the study for future research direction in this area. The SEM accepted variables are considered as the input for the next layer of ANN analysis that identied IOT has the major effect on Fog Computing (FC). A comparison is also done on SEM and neural network results. The outcome of the study will help more number of Smart City (SC) creations and will full the target of 100 Smart city creation by Government of India taking forward towards a sustainable development.

According to Wojick (2016), the new technologies like IOT, Big data analytics and Industry 4.0 will help the managers to ful l the industry needs. Although these current technologies have certain advantages, still security is a major concern in this area. Now, the time is to know what encourages people to go for Internet of Things and what will discourage them from using smart devices like smart phones (Li et al., 2016). Internet of things again is associated with people, services, energy and Fog-computing for the usage, implementation and storage of data (Lom et al., 2016).
Cloud computing has the property of virtualization that can store a big amount of data over the internet and can provide the same to the organizations and the users dynamically (Robert & Jr, 2002). Cloud IOT technology is evolved to big data analytics, and now the age is of Industry 4.0. Cloud IOT provides the required storage, networking and applications to support all the current technologies. The huge data can be minimized and stored in the fog. Fog computing can support Big-data in terms of storage and networking the smart devices (Muntone, 2013). Fog computing and IOT solutions can give required information to guide the healthcare sector (Rifkin, 2011). IOT is recently applied to medical services that increase e ciency to cure the patients' illness (Hermann, 2016). The combination of cloud computing and IOT along with

Background Study
Internet of things (IoT) can be combined as internet for people to ease the work, internet for different services, internet for energy saving. And the entire data generated from it are stored in the Fog computing (Patel & Patel, 2016). So, it becomes the powerful network of physical things through the internet without any human interference.

Internet of Things (IoT)
The upcoming revolution in Internet is known as IOT (Internet of Things). It has become a multidisciplinary research area in so many Industry, healthcare and institute speci c studies. With the inception of multiple smart devices and smart phones, the IOT sensor machines are developing e ciently based upon the needs in our day to day life. Because of IOT, the healthcare system would be called as a IoT offers device to device connectivity through some smart services known as Internet of Services (IoS). These services are provided through the sensors and smart-grids. Internet of People (IoP) can connect the people with the environment that identi es the real world and virtual world. Energy saving is a very important concept. The energy produced can be saved through remote monitoring and the actual demand of energy can be known from the data generated from IoT where Internet of Energy (IoE) comes into role

Smart-City (SC)
Smart city and IOT are popular phenomenon to resolve the nancial crisis. Smart city is not only a technical aspect; a lot of humanitarian points are also involved. Smart City brings together all the technologies to enable smart movement, smart economy, smart individuals, smart governance and smart living. It ensures the sustainability of the smart city improving the quality of life and safety of people with the latest technologies in six key areas of environment, economy, people, governance, living and mobility (Lom, 2016). A constant effort is given by the European Union (EU) for achieving the growth in its metropolitan city (Paskaleva, 2009). The revolution in cities to smart cities should be implicit through some sequence of procedures. Smart city is de nitely an important topic as in last few years as majority of people are moving from villages to cities. To become attractive and competitive, a city requires so many techniques. It does not need an innovative pattern how the smart cities should be like, and requires how the businesses run together, people and the academic thinking among them. The cities should be reinvented as the people should not be considered as the users, they should be treated as the key stakeholders. Technologies should be used as a dynamic enabler. Business should be considered as partners, which will be a real transformation. A smart product monitors the entire life cycle of call, manufacture and distribution to the endpoint ( and supply-chain. A smart city is associated with some aspects of smart environment, smart transport, smart governance, smart people, smart livelihood, and smart mobility.

Fog-Computing (FC)
Fog-computing can be used to build the smart city where the signi cant data can be directed to the upper level and the rest of the data can be utilised in the local edges. It is a distributed computing technique used to store enough amounts of data. Some applications of fog-computing are stored in smart devices and some are in cloud. Its aim is to improve e ciency by reducing the amount of data sent to cloud for processing and storage. This can also be done for security and privacy reason (Bar-Magen, 2013).
The distributed approach of fog computing gained the popularity because of IOT technology. IOT contains so much of information and can be separated out in the next level to increase the e ciency (United Nations, 2015). IOT and smart city both can be connected through fog computing.

Research Gaps
The articles discussed about Smart city, Big-data analytics, Cloud computing and Internet of Things (IoT) adoption in Industries. Some studies are Industry speci c like SMEs and multiple case studies methods. Few studies carry certain research models and some are surveys on Smart city. Rarely the discussion of Fog Computing is done for building a Smart city which has the ability to minimize the data storage. There are certain variations between cloud computing and IOT, but rarely we connect this to Fog-computing which stores the intermediate data with energy e cient. The critical success factor identi cation for the success of smart-city is done in very speci c studies. No study identi es effectiveness of variables on Smart city.

Conceptual Model & Hypotheses Development
It is the inter-relationship between the endogenous and exogenous variables in a structured manner. Some kind of dependent and independent relationships are there among the factors.
At initial stage some logical grouping is done among the items to form the constructs. Data was collected from 13 Smart-cities out of 20 smart-cities in India for pilot study. After the pilot analysis, the nal set of variables was identi ed. With the help of nal variables, the research model and hypotheses were framed. A one-point to seven-point likert scale was used for data collection. The number of items was reduced after the pilot study. Five hypotheses were framed in the study.
The following hypotheses are formulated from the above conceptual framework.

Research Methodology
Sample characteristics 390 responses were received from 13 smart-cities in India out of which 379 responses were valid. Per city wise distribution was 15%-25% in each smart-city. 50 % of the respondents were male compared to female respondents. For 38 valid items in the questionnairre, as per 1:10 ratio, maximum 380 respondents can be taken (Hinkin, 1996). So, 379 responses were identi ed as valid responses through factor analysis and considered nally for further study. More of the respondents were in the age group of 21 to 35. Male respondents were more compared to the female respondents. The undergraduate respondents were more than the post graduates. Mostly south India was the target geographical region for data collection which almost took ve months.

Exploratory factor analysis
As per Hair et al. (2010) factor analysis helps in data reduction and identi es the structure among the variables and items. In factor analysis, majorly we check the KMO (Kaiser-Meyer-Olkin) value to know the accuracy of the data. The KMO value more than 0.65 is good for further analysis. For new scale development varimax technique is used. Rotated component matrix is taken into consideration for con rmatory factor analysis in next phase.
Con rmatory factor analysis Con rmatory factor analysis determines the model t along with the validity and reliability. Reliability is measured through composite reliability (CR) and validity is measured through discriminant validity by AVE (Average Variance Extracted). The CR values more than 0.7 is acceptable and AVE should be more than 0.5. The correlation among the factors is measured by the correlation matrix. For discriminant validity the square root of AVE must be larger than the correlation coe cients in the matrix (Malhotra, 2010& Hair et al., 2010.

Structural Equation modelling
Structural equation modelling is a widely accepted technique for hypothesis testing. It is a better alternative as compared to multiple regressions. SEM can identify the error terms whereas regressions don't. There are two types of errors in SEM, one is residual error and another one is measurement error.
The standardized regression weights measure the acceptance of each hypothesis. Squared multiple correlation identi es the weights of each residual error.

Empirical Result
In statistical analysis, EFA was performed using the IBM SPSS statistics V20, and for CFA and SEM, AMOS V22.0 was used. Discriminant validity of individual variable was identi ed through squire root of AVE that must be more than the correlation coe cients (Fornell and Larcker, 1981). The composite reliability and Cronbach's alpha of the factors crosses the threshold limit 0.7. All these Composite reliability (CR), Discriminant validity (AVE) and Cronbach's alpha are shown in the following table. Figure 2 signi es the path diagram for Con rmatory factor analysis.

Structural Equation Modeling:
The model t is measured through the estimates of the model through regression co-e cient, goodness of t index (GFI), adjusted goodness of t index (AGFI), Comparative t index (CFI) and root mean square error of approximation (RMSEA). Degrees of freedom (CMIN/DF) should be below 3.0, in our case it is 1.511 which is acceptable. The CFI value is 0.976 which should be more than 0.95. GFI is 0.907 which is more than the threshold value of 0.90. AGFI is 0.894 which is more than 0.85 threshold limit. RMSEA is 0.032 which is less than 0.05, hence acceptable (Hair et al., 2010). The following gure shows the SEM path diagram.

Hypothesis testing
The ve hypotheses were tested using SEM in Amos 20.0. In this research model fog computing is the mediating variable, smart city is the dependent variable and the rest four variables are the independent ones. The analysis shows out of ve hypotheses four were supported and one is not. The hypothesis which is not supported is the IOE that positively supports fog computing having -0.022 regression weight.
Rest four hypotheses were supported as depicted in the following table. MLP (multi-layer perceptron) was used to check the hidden layer in the arti cial neural network. The three layers (input, hidden and output layers) were shown in the above gure. A data partition was done for training and testing data in 90:10 ratios for tenfold cross validation in SPSS. The predicted variables signi cances were determined through the non-zero synaptic weights. The non-compensatory ANN analysis will be able to balance the drawbacks of compensatory SEM analysis.

Conclusion
The present study identi ed six critical success factors that in uence smart-city adoption. Structural equation modelling is used for analysis. The relationship between each factor is identi ed through exploratory factor analysis. Based upon the research questions, research objectives were framed for the study. It may be noted that smart-city is controlled by fog-computing that enables the objects in the smart-city. IOT is the major technology for the establishment of a smart city. It is clearly visible from the study that fog-computing affects smart-city majorly with higher value of regression weights of 0.119. IOT, IOS and IOP have signi cant effect on fog computing. The positive effect of IOT on fog computing also explained in the previous studies (Lom et al., 2016).
In our present study IOE negatively affects fog computing, it shows Internet of energy has negligible effect for building a smart-city. Real time implementation of smart-city is possible through fog-computing and IOT. The study is useful for all IOT enabled smart cities in India and will help the academicians to do further study.

Managerial Implications & Future Scope
Different methodology can be used to check the outcome of the study. The rst managerial implication says IOT and Fog computing alone has less impact on Smart city initiatives, but combined these technologies can solve the wonder. Secondly, the integrative dual-stage model can tackle to the new business models. Thirdly, IOT will provide a mechanism to provide the data online for the customers in cloud, but it leads to the security issues as a new research agenda.
Comparative studies will be more helpful to get the correct output. It will help the managers to frame a smart city more accurately. The study will guide government agencies to frame policies for smart-city. The IOT service providers will work more effectively by showcasing the result output as their strong point.

Declarations
Con ict of interest: I declare that it is the sole work by me and there is no con ict of interest exist in my work.
Human and animal rights statement: Humans/animals are not involved in this work.