OPTIMAL SELECTION OF NETWORK IN HETEROGENEOUS ENVIRONMENT BASED ON FUZZY APPROACH

Since quality of service of the wireless network is changing over time, there is need of systematic and periodic analyzing of network traffic to get connected with the optimal network among the heterogeneous technologies. To cope with this scenario, this work has proposed a stream data mining based on ST-DR (Dynamic relaxation) fuzzy c means clustering to classify the network traffic effectively. Subsequently classified data would be sent to the web usage mining based on Probabilistic Latent Semantic Analyzer which analyzes the traffic information. Even though being classified and analyzed the network traffic information itself can’t get connected to the optimal network due to its dynamic nature so to handle this situation this work has incorporated an ensemble machine learning algorithm applying sequential AdaBoost which predicts the quality of service of the each network during network traffic and enables the user to get connected with an optimal network which have superior Quality of Service (QoS).


INTRODUCTION
The utilization of internet in modern era is increasing by leaps and bounds. Due to this explosion, we have continuous new network generation such as 3G, 4G and 5G as well as development of new mobile devices equipped with multiple network interfaces such as bluetooth, Wifi, 3G and others. These problems can be encountered by Heterogeneous Wireless Networks (HWNs).
The major advantage of these HWNs is lies on the complementary characteristics between different access networks. For that, the optimal network selection is required for providing continuous connectivity [1] even during the roaming from one network to other.
The optimal network selection problem is tedious and very challenging. Several mathematical algorithms using the utility theory, the cost function, fuzzy logic, game theory, combinatorial optimization, Markov chain etc. have been developed for proper selection of network [2], [3], [4], [5] Since topologies of heterogeneous network can changes very frequently due to its dynamic nature, it is not easy to find the optimal network and get connected. Therefore, in order to select the best optimal network in heterogeneous networks, this work has focused on designing a superintend network traffic analyzer. We proposed a stream data mining based on ST-DR (Dynamic relaxation) fuzzy c means clustering to classify the network traffic effectively.
Subsequently classified data would be sent to the web usage mining based on Probabilistic Latent Semantic Analyzer which analyzes the traffic information. Even though being classified and analyzed the network traffic information itself can't get connected to the optimal network due to its dynamic nature so to handle this situation this work has incorporated an machine learning algorithm based on sequential adaboost which predicts the quality of service of the each network during network traffic and enables the user to get connected with an optimal network. The outline of the paper encompasses Section 2 with discussion of the related researches which coordinates with the field of research, Section 3 with brief description of the proposed methodology and the working characteristics of the superintend network traffic analyzer, Section 4 gives the results and discussion of our proposed method and Section 5 provides conclusion of this paper which is then followed by references.

LITERATURE SURVEY
Some of the recent researches related to our research are given below: Santosh Kumar Das et.al [11], [13] proposed a Fuzzy based Intelligent Energy Efficient Routing protocol (FIE2R) for WANET. Mohamed Lahby et.al [12] proposed a new mechanism based on graph theory for the selection of the best path for network selection decision. Pengfei Ren et.al [14] studied a distributed machine learning problem by applying a distributed optimization algorithm over an undirected and connected communication network. Raman Kumar Goyal [15] proposed a novel fuzzy-Analytic Hierarchy Process (AHP) based network selection in heterogeneous wireless networks. Triangular fuzzy numbers are used to represent the elements in the comparison matrices for voice, video and best effort applications.
Consequently, the recent works discussed above vigorously epitomize the demand for designing a superintend network traffic analyzer to enhancequality of service (QoS) through properly analyzing the network traffic and connected with the optimal network along with the use of graphical representation. Thus, in this work it is focused to deliberately design a superintend network traffic analyzer which brings efficient results for the improvement of quality of service (QoS).

STURDY NETWORK TRAFFIC ANALYZER BASED ON FUZZY APPROACH
Wireless Sensor Networks (WSNs) are formed by hundreds or thousands of low cost and low energy sensor devices. To provide better QoS, it is necessary to analyze the traffic loads in each sensor networks and afford an optimal network for data transmission. To obtain optimal network, data mining is utilized for analyzing. Figure 1 depicts the general architecture of WSN.  Existing techniques based on statistical approaches have suffered in selecting best network due to lack of less amount of past information as well as lack of functioning in proactive manner. Therefore, in order to choose the best optimal network in heterogeneous networks based on quality of service, this work has focused on designing a superintend network traffic analyzer. Even though being classified and analyzed the network traffic information itself can't get connected to the optimal network due to its dynamic nature so to handle this situation this work has incorporated an ensemble machine learning algorithm based on sequential adaboost which predicts the quality of service of the each network during network traffic and enables the user to get connected with an optimal network which holds high packet delivery ratio, less packet loss and high throughput. This entire proposed method is represented as a diagrammatic block diagram in figure 2.

Stockwell Transform (ST)
The S-transform has been applied in order to reduce the size of the transmitted data without losing the significant features of the data.. The ST of a time series ( )  u is defined as [16] follows: The standard deviation of the opening o of the standardST in (1) is as follows: ( ) For the modified Gaussian opening, we have chosen the standard deviation Thus, an alternative representation for the generalized ST with modified Gaussian opening is as follows: The discrete version of the ST of a signal is obtained as is as follows: In (8), it can be seen that the noise can be removed from the ST output by a simple thresholding technique.

Fuzzy C Means (FCM)
Fuzzy C Means (FCM) is a method of clustering which allows one piece of data to be in the right position to two or more clusters. Starting with an initial guess and by iteratively updating the cluster centers and the membership grades for each data point, FCM iteratively moves the cluster centers to the correct place within a dataset. This iteration is based on minimizing an objective function that symbolizes the distance from any given data point to a 561 OPTIMAL SELECTION OF NETWORK cluster center weighted by that data point's membership rank. The fuzzy c-means (FCM) algorithm [17] has numerous applications. It is based on the concept of fuzzy c-partition, introduced by Ruspini [18].
In fuzzy C-means clustering, we determine the cluster center m  and the membership matrix V . It is based on the minimization of the following objective function: where a is the number of clusters,

Probabilistic Latent Semantic Analyzer
The overall process of Web usage mining consists of three phrases: data preparation and transformation, pattern discovery, and pattern analysis. The discovered patterns may then be analyzed and interpreted for use in such applications as Web personalization [16].
One of the main advantages of PLSA model in Web usage mining is that it generates probabilities which quantify relationships between Web users and tasks, as well as Web pages and tasks. From these basic probabilities, using probabilistic inference, we can derive relationships among users, among pages, and between users and pages. Thus, this framework Bayes' rule to these probabilities, we can generate the probability that a certain task is chosen given an observed user session: (11) and the probability that a certain task is chosen given an observed pageview: Here, j k refers the latent factor, p e stands for user session and q g represents pageview. The PLSA analyzes the entire network information effectively. Though, it is impossible for the system to get connected with the optimal network. In order to obtain better connection with the optimal network, we have to predict the QoS of various network and find appropriate network that have high packet delivery ratio, less packet loss and high output

Ensemble machine learning algorithm based on sequential adaboost
In machine learning, ensemble methods use multiple learning algorithms to obtain better predictive performance than could be obtained from any of the constituent learning algorithms alone [20]. Boosting involves incrementally building an ensemble by training each new model . By far, the most common implementation of boosting is Adaboost. AdaBoost training process selects only those features known to improve the predictive power of the model [21].

Training phase
AdaBoost indicates a specified method of training a boosted classifier of the form is the weak learner that is being considered for addition to the final classifier.

Weighting
At every iterative process of the training process, a weight on that sample. These weights can be used to inform the training of the weak learner, for instance, decision trees can be grown that favor splitting sets of samples with high weights. . Thus, adaboost classifier predicts the network that has better QoS and allows the user to connect with that ideal network.

Derivation
As a consequence, our proposed method effectively analyzes and classifies the optimal network for effective data transmission with superior Quality of Service (QoS). Initially the network information is preprocessed and the noise is eradicated by Stockwell Transform (ST).
Then it is send to fuzzy c means (FCM) clustering for classifying the network information.
However, the FCM clustering attains 80% accuracy only. Therefore, Dynamic Relaxation (DR) algorithm is utilized to obtain maximum accuracy. Afterwards the network information is analyzed through Probabilistic Latent Semantic Analyzer for getting the genuine information about network traffic in each sensor node. Thus, the method classifies and analyzes the network information automatically though it doesn't able to connect with the best network because of its dynamic nature. Hence, we utilize an ensemble machine learning algorithm based on sequential adaboost that forecast the network information even during the network traffic and empowers the user to connect with the best network which has better QoS. The overall architecture of this system efficiently performs the classifying and analyzing task and helps the user to attain better network for communication.

RESULT & DISCUSSION
This section provides the entire results obtained from our proposed method which was explained in detail in the above section 3. We initially perform the stockwell transform for eradicating the noise in the network information. Then fuzzy c means clustering technique is applied in the noise removed data in order to classify the network information. However, 80% of accuracy is obtained in the fcm method; hence to achieve more accuracy we utilize dynamic Though the aim of our method is to forecast the network traffic of each sensing node and empowers the user to connect with the best optimum network. Hence to accomplish this process, this method uses ensemble machine learning algorithm based on sequential adaboost, which diagnose the QoS of each network and predicts the optimal network to connect. Consequently, our proposed method attain all the above goals efficiently which is executed using graph theory and also in graphs are presented in this section.

Figure 3: Stockwell transform (ST)
In the above figure 3, ST analyzes the noisy network information from the safe data and eradicating the noise from the entire information in order to obtain noiseless network information.     Figure 7 shows the output obtained from ensemble machine learning algorithm based on sequential adaboost method that predicts the QoS of each network and helps the user to connect with ideal network.

A. Packet Delivery Ratio
The calculation of Packet Delivery Ratio (PDR) is based on the received and generated packets as recorded in the trace file. In general, PDR is defined as the ratio between the received packets by the destination and the generated packets by the source. The figure 8 shows the packet delivery ratio of proposed scheme and existing scheme. Compare to existing scheme the proposed scheme gives 98% of the packet delivery ratio.

B. Packet Loss Ratio
Packet loss happens when one or more packets of transmitting over a communication channel do not reach the destination. Packet loss is measured in terms of percentage of packets loss with respect to the packets sent. The figure 9 shows the packet lost ratio of proposed scheme and existing scheme. Compare to existing scheme the proposed scheme reduces 40% of the packet lost ratio.

CONCLUSION
In this paper, we designed a superintend network traffic analyzer for classifying and analyzing the optimal network effectively for enabling the user to get connected with the optimal network Ultimately, to predict the QoS of each and every network during network traffic, an ensemble machine learning algorithm based on sequential adaboost is utilized and enables the user to connect with the ideal network that has maximum QoS. The proposed system accomplished the entire projected performance with greater accuracy.