Mining IOT

Mining Internet of Things for intelligent objects using genetic algorithm

Abstract

The Internet of Things (IoT) is overpopulated by a large number of objects and millions of services and interactions. Therefore, the ability to search for the right object to provide a specific service is important. The merger of the IoT and social networking, the Social Inter- net of Things (SIoT), has made this possible. The main idea in the SIoT is that every object in the IoT can use its friends’ or friends-of-friends’ relationships to search for a specific service. However, this  is  usually  a  slow  process  because  each  node  (object)  is  required to manage a large number of friends. This paper addresses the issue of link selection of friends and analyzes five strategies in the literature. Then it proposes a link selection strategy using the Genetic Algorithm (GA) to find the near optimal solution. The results show an improvement over the examined strategies in terms of several parameters.

Introduction

The Internet of Things (IoT) is considered the next evolution of the current global Internet [1]. The main idea is to increase its ability to gather, analyze, and distribute data and transform them into information, knowledge, and wisdom. However, it is not about connecting people. It is about connecting things, hence its name. It covers many possible application areas, and it enables objects to connect anytime, anywhere, and to anything.

In the IoT, a thing could be anything and everything, from a mobile device or a dishwasher to a controlling system of a car or a plane. It can be absolutely anything that moves or does not move. If it has an IP address, it is possible to connect it or track it. Thus, these things are not just smart phones and tablets; they are everything [2].

The IoT includes a vast number of objects that generate information about the physical world. This information can be obtained through standard Web browsers. In addition, the IoT can provide new services to end-users. However, in [3], the authors explained that the search of each service in the IoT is huge because the number of objects that connect to the network is continuously and rapidly increasing.

In addition, the traditional interaction model is  based  on  the  idea  that  humans  are  looking  for  information  (human- object interaction). However, in the IoT, this model must change to object-object interaction, which means that an object will look for a service from other objects. In the literature, several models were proposed for real-time search [1,4]. However, these traditional models employ centralized systems for their engines; hence, they do not scale properly with the number of devices and queries. In order to overcome this shortage, a new approach based on the Social Internet of Things (SIoT) was proposed [4].

The SIoT can be used as an analog term for “social network of intelligent objects” [5]. Therefore, the SIoT can be thought of as the ability to have  integration  between  the  IoT  and  social  networks  in  an  intelligent  way  [4,6].  In  the  SIoT,  objects will have the ability to search for a  desired  service  using  its  friends’  objects  through  available  connections  between  them (i.e., friendship connections). As a result, each node will eventually have a large set of nodes (friendships) to manage, which will negatively affect the search time. Therefore, it is advisable to limit the number of friendships for each node. Moreover, choosing which friendships to keep will affect the search efficiency [7].

In the SIoT, every node is an object that can establish social relationships with other things in a predefined way, according to the rules that where set by the owner [6]. Many types of relationships exist [8]:

  1. Parent-object relationship (POR).
  2. Co-location object relationship (CLOR).
  3. Co-work object relationship (CWOR).
  4. Owner-object relationship (OOR).
  5. Social-object relationship (SOR).

This paper addresses the issue of link selection of friends and analyzes five strategies in the literature for this purpose. It then proposes and implements a link selection strategy using the Genetic Algorithm (GA) to find the near optimal solution (near optimal link selection).

The rest of the paper is organized as follows. Section 2 discusses some works that are related to this topic. Section 3 evaluates the performance of some strategies that are proposed in the literature. Section 4 includes the authors’ proposed GA for link selection, and Section 5 discusses sample performance results. Finally, Section 6 provides some conclusion notes.

Conclusion

Object search in IoT is considered an important issue due to its large and complicated search space. This complication rises from the fact that every object in the IoT can use its friends’ or friends-of-friends’ relationships to search for a specific service. The proposed Genetic algorithm based technique was introduced to overcome the limitations of some of the art of the state algorithms. The paper first discussed five heuristic search functions introduced in the literature. A new genetic algorithm based search algorithm is then introduced to find the near optimal solution (near optimal link selection).

The proposed strategy for the link selection in the SIoT achieves better results in terms of average degree and average cluster coefficient. However, the authors’ strategy gives a slight enhancement in terms of the average shortest path length.

For future work, the authors recommend designing and implementing a hybrid GA fitness function to overcome the shortcoming of the traditional  function.  In  the  suggested  function,  the  optimal  solution  would  be  the  chromosomes  with the shortest path and the maximum cluster coefficient.

No Comments

Post A Comment