Abstract
Ubiquitous computing (UC) is an advanced computing concept that makes services and computing available everywhere and anytime. In UC, data lies at the heart of all UC applications, and the key technologies that are required to make UC a reality are data and task management. In this context, retrieving data is influenced by the dynamic nature of these systems in addition to human and sensor failures. So the main problem is how to select the most appropriate service provider for retrieving data. Retrieving data is a complex issue that requires an extensive evaluation process and is one of the biggest challenges in UC. In addition, not every eventuality in these systems can be predicted due to their dynamic nature. As a result, there is a strong need to address the uncertainty in context data. In this paper, to assist users to efficiently select their most preferred service provider for retrieving data, a new fuzzy integrated multicriteria decision-making model, which meets quality of context (QoC) and quality of service (QoS) and satisfies user quality requirements and needs, is proposed. The proposed model is based on four stages. In the initial stage, the identification of evaluation criteria is performed due to the varying importance of the selected criteria. In the second stage, a fuzzy Analytical Hierarchy Process (FAHP) procedure is utilized to assign importance weights to each criterion. In the third stage, the fuzzy Technique for Order Preference by Similarity of an Ideal Solution (FTOPSIS) is used to evaluate and measure the performance of each alternative. Finally, sensitivity analysis is performed to check the robustness and the applicability of the proposed model.
1. Introduction
Currently, the human life is a world full of computing devices as shown in Figure 1. By 2025, the number of smart device subscribers will have reached 5.9 billion. Due to this increase, computers are playing an ever-increasing role in the daily lives of people. As a results, ubiquitous computing (UC) [1] can be considered as a complete view for the future that is move near to implementing at an hasten space [2].
In UC, the computing devices are embedded into the physical environment [3, 4], so the users can interact with the devices at the same time they interact with the physical environment in a highly distributed fashion. The diverse devices are linked to each other and have diverse sizes and input and output capabilities depending on their objectives, which differ in terms of hardware components, operating systems, and capability. These features of UC inspire a need for interaction methods that are radically diverse from the desktop computer interactions.
The ubiquitous computing concept has been used by several systems. They are characterized by their ability to adapt their operation to the surrounding context to improve usability and efficacy. Because of the wide spread and heterogeneity of devices in the ubiquitous computing environment, the dynamic nature of the environment, the frequent changes in user needs, and the possibility of unforeseen events during execution, these systems pose considerable challenges [2]. Indeed, these systems use a variety of devices, sensors, and networks to form a heterogeneous distributed environment, integrated into the daily activities of users.
In these systems, not every outcome can be predicted, because of their dynamic nature. As a result, there is a strong need to deal with the uncertainty in context data. Decisions based solely on the context data available can obstruct the overall system. As a result, assessing the quality of context (QoC) is a critical issue because it may compromise decision correctness and timeliness. In addition, the uncertainty in the context data might lead to poor decision-making. So deterioration of future context prediction abilities can potentially be a problem.
In this scenario, It is essential to understand the quality of information and services in order to ensure efficacy and user pleasure while also respecting the quality of context and service. As a result, the selection of acceptable services from among those offered by a number of different providers is based on the quality of contextual information, referred to as quality of context (QoC), and/or the quality of the services supplied, referred to as quality of service (QoS) [5]. Moreover, during the execution, it is necessary to ensure that the services and context information continue to meet the QoC and QoS criteria of context-aware ubiquitous systems.
In UC, a set of users may request a service from a certain service provider which needs some data items to execute the requested service for a user [5]. The required data items are stored in a set of nodes that gather the data from environments, and there are many nodes which provide data with similar functionalities with different possessing criteria. So, from where does the service provider obtain this data? And what are the criteria that determine the best node for retrieving data? Retrieving data is a complex issue that requires an extensive evaluation process and is one of the biggest challenges in UC. In addition, not every eventuality in these systems can be predicted due to their dynamic nature. As a result, there is a strong need to address the uncertainty in context data. This problem is called multi-criteria decision-making (MCDM) problem which is how to select the most appropriate service provider for retrieving data based on multiple criteria.
In this paper, to solve this problem, a new node selection model is proposed using integrated fuzzy Analytical Hierarchy Process (FAHP) and fuzzy Technique for Order Preference by Similarity of an Ideal Solution (FTOPSIS) methods to select the most appropriate node that has the required data items and satisfy the user requirements and node provider capabilities. The proposed model uses eight criteria which are mobility, waiting time, cost, trust, sharing and collaboration, workload, data transfer, and data quality of a service provider as multiple constraints and eight alternatives.
The main contributions of this paper are as follows:(1)Describing and formulating the node selection problem for retrieving data in UC.(2)Proposing a new node selection model using integrated FAHP and FTOPSIS.(3)Introducing the sensitivity analysis using applied example to check the execution of the proposed model.
The rest of this paper is ordered as follows: Section 2 introduces the related works. Section 3 describes the quality of context and service. Furthermore, it describes and formulates the node selection problem. Section 4 introduces the proposed node selection model for data retrieval. Sections 5 and 6 provide the sensitivity analysis using applied example to check the execution of the proposed model. Finally, Section 7 concludes the paper.
2. Related Work
In recent years, many schemes have been proposed for data management problems in ubiquitous computing. In the rest of this section, these schemes will be introduced in detail.
Ganchev et al. [6] presented the creation of a data management platform for use in the UCWW using a three-layer Lambda based architecture to achieve high throughput and low latency. Panigati [7] presented personalized management of semantic, dynamic data in a pervasive system (SuNDroPS), which uses context-aware approaches to collect data, provide shared services, and distribute information; the system is built on a context-aware method that, when applied to these tasks, reduces the so-called information noise by giving consumers only the information that is relevant to their current situation.
Salah and Saadi [8] developed a context-aware ubiquitous learning system, a fuzzy AHP strategy for selecting learning services. They were able to ensure a better selection of the learning service by employing the AHP technique, which allowed them to consider many evaluation factors. These requirements, on the other hand, are based on ambiguous linguistic words that fuzzy set theory should be able to describe. The fuzzy AHP approach was chosen as a result of this. The proposed approach’s methodology included identifying the learning scenario as a key stage in the selection process, resulting in a high-quality context and a high-quality service.
O’donoghue and Herbert [9] described the architecture of the Data Management System (DMS), which uses agent-based middleware to intelligently and effectively manage all pervasive medical data sources. Lee et al. [10] created a management system that is appropriate for diabetic individuals based on their blood sugar levels. Dorj et al. [11] created IHDMS (Intelligent Healthcare Data Management System) with nano-sensors for a mobile application. The proposed IHDMS organizes patient healthcare data from nano-sensors and converts it into the widely used HL7 (Health Level Seven) standard. This transformed data is sent to a system’s server.
Breitbach et al. [12] proposed an edge computing data management strategy that separates data placement from task scheduling. They introduce a multilevel scheduler that considers many context dimensions when placing data on resource providers in the system. The scheduler assigns tasks based on the current context and monitors the state during execution. The number of data copies is adjusted if necessary to optimize the trade-off between execution latencies and data management overhead.
Izadpanah et al. [13] proposed a thorough set of quality standards for ubiquitous services. This framework is designed to collect and categorize a large number of service quality characteristics. Through a poll, some specialists in ubiquitous computing assessed and changed the first proposed classification. The AHP approach is then used to rank ubiquitous services using the final recommended hierarchical structure of the QoS criterion. Xiao et al. [14] suggest a data selection approach based on data utility evaluation. To achieve dynamic data modalities selection, a customized machine learning model is trained to predict the ideal modal combination based on the quality evaluation of multimodal data samples.
In [15], an integrated design alternative assessment model integrating Z-cloud rough numbers (ZCRNs), best-worst method (BWM), and multi-attributive border approximation area comparison (MABAC) was developed to fully handle various uncertainties, reflect judgmental reliability, calculate criteria weights, and rank design alternatives.
In [16], a novel model called q-ROF-BWM-WASPAS is presented for selecting the appropriate manufacturer. This model uses the q-rung orthopair fuzzy (q-ROF) score function for measuring q-ROF values and a best-worst method (BWM) to determine the fuzzy criteria weights. In addition, an improved weighted aggregated sum product assessment (WASPAS) with q-ROF settings is presented.
In [17], an integrated assessment and prioritization method of engineering characteristics, by constructing a comprehensive list of engineering characteristic indicators, was developed. This method develops the cloud rough number model that can simultaneously handle various uncertainties, and considers the objectively optimal weight of both experts and criteria.
In [18], an applicable decision-making method to assist managers in extracting the essential remanufacturing alternatives for product improvement was developed. This method develops a new concept named dual interval rough number clouds (DIRNCs) by combining the merit of interval rough numbers (IRNs) and interval cloud model in handling uncertain information. Then, it constructs an applicable decision-making support model of remanufacturing alternatives based on DIRNCs, two nonlinear weighting methods, and a Technique for Order Preference by Similarity to Ideal Solution (TOPSIS).
Through the analysis of these works, the satisfaction of a user is not ensured because they have limitations in selection. Indeed, the majority of them have overlooked QoC evaluations and considered only QoS during the service selection process. Similarly, no work has included the identification of the server provider situation at the start of the process, which will aid in the selection phase. When it comes to the selection process, the majority of works only support the first phase, ignoring the remaining steps of reselection, evaluation, and classification. As a result, there is a need for a method for selecting service provider that takes into account the identification of situations, as well as the evaluation of QoC and QoS. By handling contextual information, this technique takes into account the quality that needs to be established by any user.
3. Conceptual Background
In this section, to understand the quality of information and services in order to ensure efficacy and user pleasure while also respecting the quality of context and service, the quality of context (QoC) and the quality of service (QoS) will be described. In addition, the node selection problem in UC will be described and formulated.
3.1. Context and Context Awareness
There are several definitions of context that have been presented in the literature. Some academics use examples to establish the context. Brown et al. [19] list various instances of context, including the location of a user, identities of those around him or her, time, season, and temperature. Similarly, Dey [20] defines context emotional state, focus of attention, position and orientation, date and time, and items and people in the user’s environment as examples of context. Ward et al. [21] refer to the status of the application environment as the context. Abowd et al. [22] provide the most widely accepted definition of context sensitivity: when a system uses context to give relevant information and/or services to a user, relevancy is determined by the user’s task. Moreover, they give the most widely accepted definition of context: any information about the service provider, its environment, and adjacent resources and objects.
3.2. Quality of Context (QoC)
It is essential to ensure the quality of the detected context information in a context-aware application. In [23], Buchholz et al. define context quality as any information that describes the quality of context information while Krause and Hochstatter [24] provide a more detailed definition. They define context quality as any inherent information that describes the context information and may be utilized to assess its worth for a particular application. The most widely accepted definition of quality of context is the following: “an indication of context compliance degree, collected by sensors for the current state in the environment and the requirements of a certain environment consumer.”
3.3. Quality of Service (QoS)
Vogel [25] uses the term QoS to describe a collection of quantitative and qualitative properties of a distributed multimedia system that enables an application to function. It is any information that describes how a service works, according to Buchholz et al. [23]. Laplace et al. [26] studied this issue and stated that QoS is defined as the relevance of the service demanded by the user and the service given. The most widely accepted definition of quality of service is as follows: “the collection of quantitative and qualitative traits and characteristics that indicate how a service functions and can suit the learner’s needs.”
3.4. Data Node Selection Problem in UC
3.4.1. Problem Definition
Assume that UC environment consists of a set of users that may request services, a set of data nodes that gather the data from the environments and store it to be used later, and a set of service provider nodes that provide the service . Assume that there is a set of criteria that represents the multi-constraint parameters that are used to rank each data node that has the required data for executing a certain service. Assume that there is a set for each user in and there is a queue of service requests , where is the number of service requests of a user .
Assume that a user in has a request for a certain service from a one of service provider nodes in . The system will select the best service provider for responding to the user request, the service provider may need some data to be able to accomplish the task, and these data are stored in many nodes in DN in the environment. So, from where the service provider obtains this data? And what are the criteria that determine the best node? This problem is called multi-criteria decision-making for data node selection (MCDM-DNS). So MCDM-DNS is how to select the most optimal data node among these data nodes to retrieve the required data items based on different criteria such that the selected data node satisfies the user requirements and node provider capabilities. Based on this description, MCDM-DNS can be defined as follows.
3.4.2. MCDM-DNS Problem
Given a set of users , a set of data nodes , , a set of criteria , and a queue of service requests for each user in , for each service request by a user , select the most appropriate data node from that has the required data items and satisfies the user requirements and node provider capabilities for executing the requested service.
4. A Node Selection Model Using Integrated FAHP and FTOPSIS for Data Retrieval in UC
4.1. Basic Idea
To solve the data retrieval problem in ubiquitous environments which is described in the previous section, a new node selection model, called integrated multi-criteria decision-making node selection model for data retrieval (IMCDM-NSDR), is proposed. The basic idea of IMCDM-NSDR is based on (1) determining the importance of each criterion in relation to the other criteria using fuzzy Analytical Hierarchy Process (FAHP) and (2) choosing the best alternative through a group of alternatives using fuzzy Technique for Order Preference by Similarity of an Ideal Solution (FTOPSIS).
4.2. The Proposed Model
In this section, IMCDM-NSDR model is proposed for selecting and ranking data nodes in UC. To treat ambiguity and uncertainty in the evaluation of data nodes in UC, this new model depends on a fuzzy framework that integrates two MCDM fuzzy techniques. These techniques are FAHP and FTOPSIS. The proposed model is based on three phases as shown in Figure 2. In phase 1, the proposed model defines the decision-making problem (DM). Then, it constructs the hierarchical structure of criteria which includes identification of criteria, alternatives, and questions by a consultant with experts and reviewing the literature. In phase 2, the proposed model evaluates and ranks data nodes using multiple steps. Firstly, the proposed model calculates the importance of the weight of node selection based on evaluation criteria using FAHP. In phase 3, it prioritizes a set of available alternatives for criteria using FTOPSIS to get the final ranking list of data nodes; then, it selects the most appropriate one of them.
4.2.1. Phase 1: Identification and Construction Phase
The steps of this phase are as follows:(1)Identification of criteria. A set of criteria are identified and proposed. From the previous research study and a chartist of ubiquitous environment, we can obtain the criteria that influence the data retrieval in ubiquitous environments: mobility, waiting time, cost, trust, sharing and collaboration, loading, data transfer, and data quality. Furthermore, a group of decision-makers (DMs) are asked to give a score for each sub-criterion.(2)Opinion and score collection. A group of responses and opinions are collected from DMs mainly for the rating of the specified criteria. The DM group consists of many experts to form linguistic data for rating FAHP and FTOPSIS.(3)Construction of a hierarchical structure. A hierarchical structure of the MCDM-DNS problem, which describes the focus of the criteria of the MCDM-DNS problem, is constructed.
4.2.2. Phase 2: Importance Determination Phase
In this phase, IMCDM-NSDR model uses fuzzy AHP to find the weights of criteria. The fuzzy AHP (FAHP) approach was developed by incorporating fuzzy set theory into the classic AHP, where each pairwise comparison decision is represented as a fuzzy number characterized by a membership function as shown in Table 1.
The ability to combine human heuristics into computer-assisted decision-making and represent human thinking and interpretation by establishing mathematical rules to work with numerical data and linguistic terms, which are easier to understand for human reasoning, is one of the main reasons for using fuzzy methods. When it comes to fuzzy AHP, there are a few options. The most often used methods for computing the relative weights of a criterion are geometric means, provided by Buckley [27], and extent analysis methods, proposed by Chang [28]. We use the geometric technique in this research, which has been shown to work in a variety of situations [29–31].
The steps of FAHP can be ordered as follows:(1)For calculating each variable’s fuzzy weights, the following formula was used:(2)To calculate , the fuzzy summation of m extent values for a particular matrix, use the following formula:(3)Further, can be obtained via summation of ; the proposed numerical values include the following:(4)The inverse of the vector can be computed, such as dividing each summation by one:
4.2.3. Phase 3: Selection Phase
In this phase, IMCDM-NSDR model uses FTOPSIS to select the optimal alternative after obtaining the best weight. Hwang and Yoon [32] were the first to design TOPSIS (Technique for Order Performance by Similarity to Ideal Solution), one of the fundamental multi-criteria decision-making methodologies. It was developed to identify solutions from a finite set of alternatives. Its underlying logic is to define the positive ideal solution and negative ideal solution. It is based on the idea that the preferred option should be the closest to the positive ideal solution (PIS) and the furthest from the negative ideal solution (NIS). The PIS is a solution that maximizes benefit criteria while minimizing cost criteria, whereas the NIS follows the reverse logic, maximizing cost while minimizing benefit criteria [33]. The TOPSIS approach takes into account the distances to both the PIS and the NIS at the same time. Finally, the optimal solution that is closest to the PIS and furthest from the NIS is found.
Main limitation of TOPSIS method is inability to capture ambiguity or inaccuracy inherent to a group decision-making situation [34]. Fuzzy theory combined with the TOPSIS method could be a solution to this problem, allowing decision-makers to incorporate incomplete and nonquantifiable information. The fuzzy TOPSIS (FTOPSIS) approach was created by Chen [35]. This approach can be used to deal with uncertainty data, using subjectivity criteria and weights to determine the most subjective alternative(s) out of feasible alternatives. The goal of this section is to introduce some fundamental principles of FTOPSIS-derived MCDMs in a simple manner.
The FTOPSIS method generally consists of seven main steps which are described as follows:(1)Compute the fuzzy decision matrix (D):(2)To normalize the fuzzy decision matrix (R), do the following: To put the multiple criteria scales into a comparable scale, the raw data is normalized using linear scale transformation. The flowing equation is used to calculate the normalized value : where(3)Calculate the normalized weighted matrix (V). By multiplying the weights of evaluation criteria with the normalized fuzzy decision matrix , the weighted normalized matrix V for criteria is obtained. where(4)Calculate the fuzzy positive ideal solution (FPIS) as well as the fuzzy negative ideal solution (FNIS) of the options as follows:(5)Calculate the distance between each alternative using FPIS and FNIS. The distance between the FPIS and the FNIS of each weighted alternative is calculated as follows: The distance between two fuzzy integers a and b is measured by . The distance between them is computed as follows:(6)Calculate each alternative’s proximity coefficient . The distances to the fuzzy positive ideal solution and the fuzzy negative ideal solution are denoted by the proximity coefficient . Each alternative’s proximity coefficient is computed as follows:(7)Sort the alternatives in order of preference. In this stage, the various alternatives are sorted in decreasing order based on the proximity coefficient . The ideal option is the one that is closest to the FPIS and the furthest away from the FNIS.
Based on the previous phases, the proposed IMCDM-NSDR illustrated in Figure 2 can select the most appropriate node for data retrieval efficiently.
5. Model Analysis and Results
5.1. Case Study of Node Selection for Data Retrieval in UC
In this section, a case study of node selection for data retrieval in UC is conducted to demonstrate the efficacy of the proposed IMCDM-NSDR model.
5.1.1. Applying Steps of Identification and Construction Phase
Based on the steps of identification and construction phase, a questionnaire with data selection criteria was produced. The criteria were derived after a review of the current literature by a number of researchers. The criteria were mainly mobility, waiting time, cost, trust, loading, data transfer, data quality, and sharing and collaboration.
The default values for the criteria were taken from a combination of previous studies and user preferences. Table 2 shows the fuzzy pairwise comparison weights of main criteria based on the opinions of DM experts. Then, the hierarchical structure of the MCDM-DNS problem in UC was constructed. Figure 3 shows the hierarchical structure of the MCDM-DNS problem in UC.
5.1.2. Applying Steps of Importance Determination Phase
First, we want to determine the importance of each criterion in relation to the others based on the initial weighting values in Table 2. This process is described as follows: Step 1: from Table 2, the following formula is used to calculate the fuzzy geometric mean of each variable in relation to the main goal: , where is the number of criteria. Step 2: for calculating each variable’s fuzzy weights, the following formula was used: And then by dividing each summation by one, the inverse of the vector may be obtained. Following those calculations, the fuzzy geometric means are expressed in the Table 3. Step 3: the gravitational center is offered. The formulas introduced above were used to calculate fuzzy weights. Furthermore, we must offer a center of gravity function in order to calculate the final weights of the fuzzy analytic hierarchy process, where lower, medium, and upper values are summed up and divided by three: .
The findings suggest that among the data retrieval selection criteria, data transmission is the most relevant element, with importance percentage of 17.8%, which is the highest value among the others. This result demonstrates that data transfer has the most competitive advantage among others. Secondly, waiting time has importance percentage of 16.6%, mobility 16%, s & c 12%, and cost 10.6%. Finally, it has been observed that loading (10.2%), data quality (9.3%), and trust (7.4%) are the least important criteria as shown in Table 4.
5.1.3. Applying Steps of Selection Phase
In order to choose the optimal location for data retrieval, the weights were determined using a fuzzy analytic hierarchy technique, as stated in the methodology. However, we have offered the FTOPSIS approach in this part to discover the optimal alternative. FTOPSIS is a multi-criteria decision-making analysis method to resolve exceedingly complicated problems by choosing the option that is the most close to the ideal.
Table 5 shows the strategies of the group of alternatives (8) that contain the data required to implement the service for users, taking into account the priorities of each alternative.
After we normalized the fuzzy decision matrix and computed the weighted normalized fuzzy decision matrix, we calculated the FPIS and FNIS. Determining the fuzzy positive ideal (FPIS) and negative ideal solutions (FNIS) from the i-th alternative was achieved by applying equations as shown in Table 6. Using the equations for the Euclidean distances in Table 7, we determined the relative equilateral triangular fuzzy TOPSIS distances to the ideal solution of each option, FPIS and FNIS.
The relative closeness coefficients are determined, given in Table 7. The alternatives are ranked in Table 7 accordingly, and we get . The best alternative is because of the space limitation.
6. Sensitivity Analysis
The sensitivity analysis is performed for checking the robustness of the proposed IMCDM-NSDR model in different scenarios. In addition, it is performed for evaluating the proposed IMCDM-NSDR model for the problem of rank reversal. If ranking model gives nonoptimal ranks in case of addition and removal of data server nodes, the model cannot solve rank reversal problem. So the proposed IMCDM-NSDR model is verified for rank reversal problem through sensitivity analysis. Here, two strategies of sensitivity analysis are studied: The first strategy is changing the number of data server nodes (DSNs) by removing existing nodes and adding new nodes. The second strategy is changing the criteria weights. In the rest of this section, the sensitivity analysis of the proposed IMCDM-NSDR model will be evaluated based on these two strategies.
6.1. First Strategy
For changing the number of data server nodes in evaluation strategy, two scenarios are possible: (1) some data server nodes are removed, and (2) some data server nodes are added in existing data server node set.
6.1.1. Removing Data Server Nodes Scenario
In this case scenario, there are 10 data server nodes: dn1, dn2, dn3, dn4, dn5, dn6, dn7, dn8, dn9, and dn10. In the first experiment, all data server nodes were considered. In the second experiment, the sensitivity analysis was performed by removing dn10. In the third experiment, dn9 and dn10 were removed. In the fourth experiment, dn8, dn9, and dn10 were removed. The evaluation result values were computed and are shown in Table 8. The ranks of DSNs are shown in Figure 4. It can be deduced from the above two results that the ranks of DSNs are consistent. Hence, IMCDM-NSDR model is robust in the scenario where DSNs are removed in experiments.
6.1.2. Adding Data Server Nodes Scenario
The sensitivity analysis for the second case scenario was performed starting with 10 data server nodes: dn1, dn2, dn3, dn4, dn5, dn6, dn7, dn8, dn9, and dn10. In subsequent experiments, three data server nodes were added: dn11 in the second experiment; dn11 and dn12 in the third; and dn11, dn12, and dn13 in the fourth. The evaluation results and ranks of all DSNs in each experiment were computed. The evaluation result is shown in Table 9, and the ranking is shown in Figure 5. It can be observed from Table 9 that the evaluation result of DSNs is consistent in each experiment. The results show that IMCDM-NSDR model is also consistent and robust in DNS addition scenario for rank reversal problem.
6.2. Second Strategy: Changing Criteria Weights
To measure the impact of criteria weights on the selection of the appropriate DSNs, the conducted sensitivity analysis is illustrated in Table 10. The objective, as suggested in several related researches [36–39], is to investigate the sensitivity of the final decision to small variations in the criteria weights during the comparison process. It is performed by slightly changing the weight values and observing the influence on the decision. Thus, six experiments were conducted, ExpW-1, ExpW-2, ExpW-3, ExpW-4, ExpW-5, and ExpW-6, as shown in Table 10. Tables 11 and 12 present the evaluation results and rank result of DNSs in these six experiments, respectively. In addition, the graphical representations of these experiment results are shown in Figure 6.
The results in Tables 11 and 12 and Figure 6 show that IMCDM-NSDR model can adapt its selection based on changing weights of criteria and it is very sensitive to slight change in their weights.
6.3. Result Validation
The proposed model, IMCDM-NSDR, has been simulated and compared with existing literature, FAHP [40] and fuzzy TOPSIS [35]. The obtained rank for each method is shown in Figure 7. It can be observed that IMCDM-NSDR model gives similar results to those given by the FTOPSIS method. It can be deduced from the results that the proposed IMCDM-NSDR is consistent for DNSs.
In addition, IMCDM-NSDR has been simulated and compared with existing literature, FAHP [41] and FTOPSIS [33], for changing weights scenarios. The rank result and its graphical representation of FAHP and FTOPSIS are shown in Table 13, Figure 8, Table 14, and Figure 9, respectively. By comparing the results of FAHP and FTOPSIS in Figures 8 and 9 with the obtained result of IMCDM-NSDR in Figure 6, we can note that the sensitivity of FAHP to changing weights of criteria is very low and the sensitivity of FTOPSIS is moderate. On the other hand, IMCDM-NSDR has very high sensitivity to the changing weights of criteria. So IMCDM-NSDR is more consistent and robust than FAHP and FTOPSIS.
7. Conclusion
In this paper, a new node selection scheme in ubiquitous environments, called integrated multi-criteria decision-making for data retrieval (IMCDM-NSDR), has been introduced using integrated fuzzy AHP and fuzzy TOPSIS methods. IMCDM-NSDR can select the most appropriate node that has the required data items and satisfies the user requirements and node provider capabilities. IMCDM-NSDR selects the best data node among the available data server nodes based on the criteria that influence the data retrieval in ubiquitous environments, that is, mobility, waiting time, cost, trust, sharing and collaboration, loading, data transfer, and data quality. The sensitivity analysis results show that IMCDM-NSDR is an efficient model. Some of limited issues in the current work are the collaboration among data server nodes and evaluating the proposed model using real scenarios and promised simulators which did not applied. So, in the future work, these issues will be considered to improve the node selection process.
Data Availability
The data used to support the findings of this study are available with the corresponding author and can be requested from him.
Conflicts of Interest
The authors declare that they have no conflicts of interest.