BIVALENCE FUZZIFIED DECISION STUMP BOOTSTRAP AGGREGATION FOR ENERGY AND COST-EFFICIENT 6G COMMUNICATION Robbi Rahim Department of Management Informatics, Sekolah Tinggi

Future Sixth generation (6G) wireless networks are anticipatedto offer entirecoverage, improved spectral, energyandcost-efficient communication.The 6G will enable a network collectivelyand offer seamless wireless connectionsbetween the devices. While the deployment of 5G is ongoing, mobile communication networks are still suffering many basic challenges such as high-energy consumption and operating costs. To address these issues, it is very important to consider and develop new technologies in next-generation mobile communication, namely 6G. Novel machine learning can potentially assist the 6G to obtain better communication. Bivalence Fuzzified Decision Stump Bootstrap Aggregating (BFDSBA) model is introduced for energy and cost efficient communication. The BFDSBA model considers the nodes i.e. devices in the forecasting process before the data communication in the 6G network. The Bootstrap Aggregative technique utilizes set of weak learners as Bivalence Fuzzified Decision Stump. For each device in the network, energy, signal strength, and bandwidth is measured. Based on the estimated resources, efficient devices are selected for the 6G network architectural design. This in turn helps to improvedata communication with lesser cost in6G networks. The result exposesimprovement of BFDSBA model than the conventional methods.


Introduction
With the development and forthcoming sixth-generation (6G), the expectationand development of the networkhave attracted a large deal of consideration.The upcoming 6G network is alargely connected complex network that able to provide users' required services with better resource utilization such as energy and cost. To attain these requirements, 6G networks design requires a noveltechnology that offers reliable andlow latency communication for many applications.

For energy and cost-efficient communication, A Multivariate Regressive Deep Stochastic
Artificial Structure Learning (MRDSASL) method was presentedin [1]. But designed technique failed to analyze the bandwidth to achieve higher delivery rate and minimizes the cost. A hybrid Finally, an extensive simulation is conducted with various parameters to highlight improvement of BFDSBA over conventional techniques.
Work is systematized into different sections. Section 2, discusses the BFDSBA model. In section 3, Simulation settings are provided. Performance evaluation and results of the proposal and conventional methods are carried out and are discussed in Section 4. Section 5 concludes the article.

Methodology
With the continuous development of technology, wireless networks helped the users to send sensitive information without any human interaction. These technologies help to improve the communication speed of the data from one device to another. In general, many wireless technologies are available in the marketplace such as 1G, 2G, 3G, 4G, 5G, and so on. These technologies diverged from one to another based on the feature values such as availability, range, performance and coverage, bandwidth, speed of data transfer, latency, and so on. The 1G (1st generation) data rate is 2.4kbps and it suffers lot of difficulties like poor network speed and it has no security. The 2G generation was introduced in 1990s with a network speed upto 64kbps. The 3G generation was established with a transmission rate of upto 2Mbps. This network combines high-speed mobile access to services with Internet Protocol (IP). Next, 4G technology was introduced with a speed of upto 20 Mbps. With the development in the user's demand, 4G technology gets replaced with 5G technology which addresses the different challenges, namely higher data rate >1Gbps, lesser latency, high device connectivity, minimal cost, and reliable quality of service provisioning. 5G wireless networks offer majoradvantages beyond LTE (i.e. 4G technology), but may not capable to meet the reliable data connectivity demands of the futuredigital world. Therefore, a novel model of wireless communication, 6G system, will developnew attractive features namelygreater system capacity, greater data rate, less latency, and enhanced quality of service (QoS) than 5G wireless networks.
6G technology is considered to be inexpensive and fast network speed data ranges up-to 11 Gbps. The most significant advantages for 6G wireless networks are the abilityto managehuge data and offers high-data-rate connectivity.
The future sixth-generation networks have the ability to support novel and various services with completely different features and requirements than the 5G network. To build an intelligent 6G network, every node must possess adequate communication, computing, and caching resources to handle intelligent operations for proving various services. The service aware architecture is shown in figure 1.  Figure 1 shows the service-aware architecture of the 6G network. The 6G will offer a virtual connection between the terminal, base stations, and centralized network. The base stations route the data service requests to centralized network.Centralized network offers the requested services to terminal namely self-driving cars, mobile devices, laptops, and so on through wireless connection with higher data rate and low latency. The 6G network will support the Reliable Low Latency Communication. It will also support security and long-distance networking, multimedia video and high-speed Internet connectivity telecommunication, navigation, and so on. A few concerns of 6G network communication is carried out to provide energy and cost-efficient communication. Since highenergy consumption increases the operating costs. To forecast energy and cost-aware communication, a novel machine learning concept called BFDSBAmodel is introduced in the 6G network architecture to scrutinize devices (i.e. node) status for enhancingreliable communication. The Bootstrap aggregating is an ensemble meta-algorithm that provides improved classification performance than any of the weak learners alone. A weak learner is a machine learning algorithm that provides the classification outcomes with the probability of some error. In contrast, a Bootstrap aggregating is a strong learner that correctly provides an improved Bootstrap aggregating performance with lesser error. For each node in the network, three features are evaluated such as energy, signal strength, and bandwidth. The energy level of the node is measured as the product of power together with time. Therefore, the energy is measured as given below, Where denotes the energy of the nodes in the 6G network, indicates power and represents the time. The energy is measured inthe unit ofa joule. Next, the signal strength of the node is measured as given below,

Figure 4 block diagram of Bootstrap aggregating
Where, denotes a received signal strength, denotes a transmitted signal power, , are transmitter and receiver antenna gain, h 2 , h 2 indicates transmitter and receiver antenna height, is distance among transmitter and receiver antenna. Received signal power is measured in decibel milliwatts (dBm).
Bandwidth is calculated as maximum rate at which data is transferred in network. It is defined as the data that sent in a given time over a particular connection.
Where, denotes a bandwidth and measured inbits per second (bps). By applying the Bivalence Fuzzified Decision Stump, the root node verifies that the estimated value of the node is higher than the threshold value using the fuzzy rule. The fuzzy rule is used to link the inputs (i.e. estimated values) with the outputs. These rules are formulated using (condition) and ℎ (conclusion).
Where, denotes anoutput of weak learner, isthreshold for energy, is threshold for signal strength, denotes threshold for energy, denotes threshold for signal strength, indicates threshold for bandwidth. Decision stump-based classification is performed as shown in figure 5.

Figure 5 Bivalence Fuzzified Decision Stump based classification
The weak learner has some training errors during the classification process. In order to improve the classification accuracy, all weak learner results are combined to attain the strong classification as follows, Where, indicates an output of ensemble classification, indicates the weak learner. After combing the weak learner, the voting method is applied to the weak learner results for accurately finding the efficient nodes. The majority votes of the weak learner are given below, Where indicates a majority votes whose decision is identified in the ℎ classifier, ' max' indicates an argument of the maximum function that helps to find the majority votes of results

Simulation results and discussions
The performance analysis of three different methods namely BFDSBAmodel and existing MRDSASL [1] and hybrid NOMA [2] are described with different metrics.

Impact of energy consumption
Energy consumption is measured as amount of energy taken by nodes to transmit the data packets. The energy consumption is mathematically estimated as given below, Where, denotes an energy consumption and it is measured in the unit of joule (J).  50  20  23  25  100  22  26  28  150  25  29  33  200  28  32  34  250  30  35  38  300  33  38  41  350  36  40  42  400  38  41  44  450  41  45  47  500 43 47 50 Table 2 demonstrates the performance of energy consumption of nodes that delivered data packets to destination. The estimated results confirm that the BFDSBAtechnique minimizes the energy consumption as compared to MRDSASL [1] and hybrid NOMA [2]. For each method, ten runs are carried out with different numbers of nodes. The performance of BFDSBA is compared to the other two methods. The comparison results prove that the energy consumption of the BFDSBA model is considerably minimized by 12%, and 18% when compared to MRDSASL [1] and hybrid NOMA [2]. consumption for all three methods gets enhanced. Among three methods, proposed BFDSBA minimizes overall energy consumption. This is due to BFDSBAuses the Bivalence Fuzzified Decision Stump Bootstrap Aggregating model discovers energy-efficient nodes to participatein data transmission process. This minimizes overall energy consumption during the data transmission from one to another.
Observed results indicate that the BFDSBA model achieves greater packet delivery ratio than other two models. Let us consider 30 data packets sent. By applying the BFDSBA model, 28 data packets are send and delivery ratio is 93%. The number of packets delivered 27 and 26 data packets are successfully send and delivery ratio of the MRDSASL [1] and hybrid NOMA [2] are 90%, 87%.
The average of comparison results inthe BFDSBA model enhances packet delivery ratio by 6% and 10% than the [1] and [2]. Figure 7 illustrates results of packet delivery ratio forthree models. Observed results prove that the BFDSBA model achieves a higher data delivery ratio. The BFDSBA model finds the higher energy, signal strength, and bandwidth nodes to carry out communication. This assists to enhance data transmission and reduces packet drop.

Impact of cost
Cost is a metric measured in terms of delay during the data transmission. The delay is measured as difference amongtime for data packet arrival and transmitting from node. It s mathematically formulated as given below, Where, denotes a data arrival time, indicates the transmission time. Delay is calculated in milliseconds (ms).

Conclusion
A novel machine learning technique called BFDSBA is introduced in this paper to forecast the energy and cost-aware communication in 6G network. BFDSBA uses the Bootstrap Aggregating ensemble technique to find the efficient devices for communication by constructing the weak learners. The Bivalence Fuzzified Decision Stump is a tree to analyze the energy level, signal strength, and bandwidth using fuzzy conditions and returns the output. The nodes that have better resources are considered to perform the data communication in the future 6G networks. It will help to improve reliable data communication with lesser cost in6G network. Assessment results demonstrateefficiency of BFDSBA model is better in terms of delivery ratio, energy consumption and cost.

Conflict of Interests
On behalf of all authors, the corresponding author states that there is no conflict of interest.

Data Availability statement:
Data sharing not applicable to this article as no datasets were generated or analysed during the current study.

Code Availability
Not Applicable.