This is an open access article under the CC BY license : Al-Khwarizmi Engineering Journal Al-Khwarizmi Engineering Journal ISSN (printed): 1818 – 1171, ISSN (online): 2312 – 0789 Vol. 21, No. 4, December, (2025), pp. 82-92 Performance Evaluation and Optimisation of Spectrum Management in Communication Systems by GA Ahmed A. Thabit 1*, Bassam H. Abd 2, Ali J. Al-Askery 3 and Haider W. Oleiwi4 Electrical Engineering Technical College, Middle Technical University, Baghdad, Iraq 1,3 2Electrical Engineering College, University of Technology, Baghdad, Iraq 4Department of Electronic and Electrical Engineering, Brunel University London, UK *Corresponding Author’s Email: ahmed.thabit@mtu.edu.iq (Received 12 June 2025; Revised 7 October 2025; Accepted 27 October 2025; Published 1 December 2025) https://doi.org/10.22153/kej.2025.10.002 Abstract Whilst wireless communication technologies proliferate, putting extra demand on the finite radio frequency spectrum and leading to issues of congestion, underutilisation and interference, this dissertation presents a modern spectrum management model on the binary genetic algorithm (BGA) capable of improving detection accuracy and adaptive spectrum access in cognitive radio networks (CRNs). BGA follows binary encoding to determine optimum weighting factors for secondary users in a CRN scenario with a much faster performance and better reliability than conventional genetic approaches. In the cooperative spectrum-sensing scheme proposed in this paper, multiple secondary users will forward their local sensing outcomes to a fusion centre in which BGA optimisation will fine-tune the weighting coefficients throughout the soft decision fusion mechanism. The algorithm then evolves from one generation to the next through the application of selection, crossover and mutation operations to discover the best configuration. Extensive simulation experiments were conducted to study the effects of the critical genetic parameters of mutation probability, crossover rate and population size on detection capability. The results indicate that the optimised BGA framework can achieve detection probability close to 96%, false alarm rate of 0.1, mutation rate of 0.12 and bit error rate of around 7 × 10⁻⁵ even when the signal-to-noise ratio is extremely low at –15 dB. In addition, the comparative evaluation showed the definite superiority of the proposed algorithm when tested against conventional algorithms, such as energy detection, matched filtering and neural network-based convolutions, when subjected to challenging and noise-prone conditions. The work further affirms the applicability of evolutionary algorithms in enhancing the cognitive intelligence of CRs and presents a scalable solution for spectrum management in existing 5G systems and future 6G frameworks. Keywords: Cognitive Radio; Crossover; Detection Rate; Dynamic Spectrum Access; Genetic Algorithm; Optimisation; Spectrum Sensing 1. Introduction Considerable changes in social automation and connectivity have been brought about by groundbreaking advances in technology related to wireless communication infrastructures. Changes in this area are responsible for the exponential increase in the number of connected devices and user demands, thereby complicating and aggravating problems of spectrum congestion and the efficient use of limited frequency resources. Additionally, all these limitations affect the overall performance and quality of service of modern-day wireless networks [1,2]. The modern generation of wireless technology is evolving to meet the insatiable appetite for ever- increasing data transfer rates along with low latencies and the efficient usage of the available spectrum as well as 3G rendering mobile multimedia capabilities, thus establishing 2G as digital voice communications. Advanced methods such as LTE, OFDM and MIMO were utilised by 4G implementations to provide services for data at high speeds. Although modern-day 5G networks mailto:ahmed.thabit@mtu.edu.iq https://doi.org/10.22153/kej.2025.10.002 Ahmed A. Thabit Al-Khwarizmi Engineering Journal, Vol. 21, No.4, pp. 82- 92 (2025) 83 offer technical advantages and opportunities, they also came with problems regarding propagation caused by its very wide frequency ranges: under 1 GHz, mid-band (1-6 GHz) and even millimetre- wave spectra [2,3]. The frequency allocation implemented in the 5G implementation strategy is illustrated in Fig. 1. Fig. 1. 5G-based electromagnetic spectral part [3] Spectral crunch refers to the limitation of a static system employing spectrum allocation with the power of current extreme growth in mobile connectivity. Therefore, DSA technologies play an important role in addressing that domain. In particular, they enable wireless devices to intelligently exploit available frequency bands depending on real-time changes in the environment. This opportunistic access of DSA considerably enhances spectral efficiency through the fair treatment of different users and minimum interference. A few characteristic application scenarios of DSA implementation emphasising its important role in the prevalent modern wireless architecture are shown in Fig. 2 [4,5]. Fig. 2. DSA applications CR innovation will be most instrumental in promoting DSA applications that enable radios to detect, characterise and adapt to the spectrum conditions. CR devices are capable of detecting unutilised frequency bands, also known as spectrum holes or white spaces, to access them dynamically without affecting any primary licensed user. This facility draws best applicability in densely populated metropolitan areas, where spectra congestion is at its worst [6]. The facilities and capabilities of telecommunication systems have witnessed dramatic changes via the AI and ML in the application of predicting traffic patterns and adjustment of sensing algorithms, thus improving performance. Past studies have shown that the next advanced generation systems networks (i.e. 6G) allow new applications with different requirements in the dynamic allocation of resources [7, 8]. Achieving high performance with good quality requires strong optimisation algorithms that are capable of managing nonlinear and dynamic situations. Advancements in 6G devices and the application of terahertz frequency range may sound futuristic, but it could also lead to greater complexities in the distribution and management of the spectrum. This makes genetic algorithms (Gas) one of the more promising bio-inspired methods for improving precision in spectrum sensing and resource allocation within cognitive radio networks (CRNs) [9–11]. In this paper, we present an improved GA-based detection system to enhance detection performance and spectrum utilisation efficiency. The effectiveness of the proposed method is tested against earlier established techniques, which appeared in various papers, via simulative comparison with the aim of establishing its superiority in the stage of variety of wireless medium. The particular contributions of this work are as follows: 1) Formulating a genetic algorithm-based joint detection model that enhances the accuracy of smart selection fusion methodologies; 2) Using a fitness-based choice, multipoint crossing and genetic variation to improve on the weighting vector for secondary users (SUs); and 3) Testing performance by extensive simulations, thus ensuring good detection reliability even under low SNR. The tuning of genetic factors, such as population, mutation and crossover frequency, will have tremendous impacts on overall performance in demonstrating GA’s capability for real-time CR optimisation. Ahmed A. Thabit Al-Khwarizmi Engineering Journal, Vol. 21, No.4, pp. 82- 92 (2025) 84 The remaining parts of this work are organised as follows: Section 2 presents a brief overview of GA with CRN, Section 3 discusses the details related to the introduced system model, Section 4 talks about the simulation outcomes, Section 5 talks about the comparison of the proposal with the current state of the art and finally, Section 6 ends the paper by presenting the conclusions. 2. Background GAs are a category of biologically inspired techniques for optimisation based on the principles underlying biological evolution. Because of their versatility and robustness, GAs are widely used to deal with difficult, multiple optimisation problems. In today’s mobile communication systems, GAs have also been remarkably effective in optimising RF utilisation. A GA starts off with a heterogeneous population of possible solutions, each represented as a chromosome. The actions are then evaluated by a fitness function as the measure of the efficiencies of each in dealing with the problem at hand. Next, this population is subjected to several generational successions under the operations of selection, crossover and mutation, similar to actual genetic processes [12–14]. Through succeeding generations, the population approaches optimal or near-optimal solutions by preserving and enhancing the most fit people. In wireless systems, genetic algorithms have shown significant potential in numerous critical domains, including spectrum allocation, which involves the successful allocation of frequency bands to reduce interference whilst satisfying the demands of diverse users. CRNs have an enhanced precision in identifying main users (PUs) and selecting suitable transmission bands based on environmental circumstances [15]. Power control and spectrum sharing involves the optimisation of transmission power and resource allocation procedures to enhance reliability and mitigate cochannel interference. Although successful, Gas are sensitive to specific design factors, such as mutation rate, crossover frequency and the number of individuals, thus requiring meticulous adjustment to guarantee convergence and prevent premature stagnation [16]. In signal processing, GAs have demonstrated promising outcomes in complex situations involving multiple-input multiple-output (MIMO) systems and intricate modulation schemes (e.g. M- PSK), by effectively manoeuvering through noisy signal environments and improving detection performance [17–19]. In conclusion, GAs offer a versatile and effective framework for dynamic spectrum management, especially in future-oriented communication networks in which real-time improvement is critical [20]. 3. System Model The significant role of sensing stage in the CRN systems is highly affected due to bad environments (e.g., fading and additive white Gaussian noise (AWGN)) in communication channels. The soft decision fusion (SDF)-based cooperative scheme is employed so as to improve detection reliability. The architecture consists of multiple SUs, so-called relays, forwarding sensed information to a central fusion centre (FC). The redesigned cooperative system model with updated notations is shown in Fig. 2 [21–25]. 3.1. Primary-relay channel modelling Each secondary user performs local spectrum sensing to detect the presence of a primary user (PU). The sensing process is formulated as a binary hypothesis test: Under the null hypothesis H₀ (no PU signal): rk[m] = ηk[m], …(1) Under alternative hypothesis H₁ (PU present): rk[m] = αk·s[m] + ηk[m], …(2) where rk[m] denotes the m-th received signal sample at the k-th SU. The variable ηk[m] represents the AWGN at SU-k, assumed to have zero mean with variance σ²ηₖ. In addition, αk denotes the sensing channel gain between the primary user and the kth SU, incorporating effects, such as shadowing and multipath fading. The PU’s transmitted signal s[m] is modelled as a zero-mean Gaussian random process with variance σ²s. Table 1 provides a reference for the variables used in Fig. 3, along with their corresponding definitions, as described in the system model. Ahmed A. Thabit Al-Khwarizmi Engineering Journal, Vol. 21, No.4, pp. 82- 92 (2025) 85 Table 1, Variable definitions used Fig. 3. Cooperative sensing system model using updated variable notations. 3.2. Relay-fusion channel modelling This subsection describes the transmission from SU to FC as seen in Equation 3: yk[m]= τk·βk·rk[m] + ξk[m], …(3) where τk is the transmission power of SU-k, βk is the amplitude gain of the reporting channel between SU-k and the FC and ξk[m] is the AWGN at the FC from SU-k, as well as a zero-mean variance of σ²ξₖ. Substituting the hypotheses into (3), under H₀: yk[m] = τk·βk·ηk[m] + ξk[m], …(4) yk[m] = τk·βk·αk·s[m] + τk·βk·ηk[m] + ξk[m] …(5) However, under H₁: Zk = Σ{m=1}N |yk[m]|². …(6) Let Zk denote the energy estimate computed at the FC from SU-k’s signal: This test statistic Zk approximates a Gaussian distribution via the central limit theorem. Its statistical parameters under each hypothesis are given as follows, where μ₀ₖ and μ₁ₖ are the expected values of Zk under H₀ and H₁, respectively: μ₀ₖ = N·(τk²·βk²·σ²ηₖ + σ²ξₖ), …(7) μ₁ₖ = N·(τk²·βk²·αk²·σ²s) + μ₀ₖ. …(8) The global test statistic Z at the fusion centre is a weighted linear combination of all individual energy statistics: Z = Σ{k=1}K λk·Zk, …(9) where λk is the weight assigned to SU-k, satisfying the normalisation constraint Σ λk² = 1. The performance is characterised using the Q-function: Pf = Q((β - μ₀) / σ₀), …(10) Pd = Q((β - μ₁) / σ₁). …(11) Each SU forwards its observations to the FC through an orthogonal reporting channel. The received signal at the FC from the k-th SU is given by: where β is the detection threshold at the FC and (σ₀, σ₁) are the standard deviations of Z under H₀ and H₁, respectively. 3.3. Conventional weighting strategies in SDF The following classical methods are used to define the weighting coefficients λk: - Equal gain combining (EGC): λk=1/√K, assigning equal weight to all SUs. - Maximal ratio combining (MRC): λk ∝ SNRk, where SNRk is the signal-to-noise ratio at SU-k. - Normal deflection coefficient (NDC): λ ∝ Σ₀⁻¹·θ, where θ = μ₁ - μ₀. - Modified deflection coefficient (MDC): λ∝Σ₁⁻¹·θ, using covariance under H₁ instead. These strategies aim to enhance detection probability by emphasising more reliable SU measurements in the fusion process. Variable Definition s[m] Signal transmitted by the primary user (PU) αₖ Channel gain between PU and the k- th Secondary User (SUₖ) rₖ[m] Received signal at the k-th secondary user (SU) ηₖ[m] Noise at the k-th SU (AWGN) τₖ Transmit power of SUₖ βₖ Channel gain between SUₖ and fusion centre (FC) ξₖ[m] Noise at the FC from SUₖ yₖ[m] Signal received at FC from SUₖ Zₖ Energy detection result of SUₖ at the FC λₖ Weighting coefficient assigned to SUₖ Z Final global test statistic at the FC Ahmed A. Thabit Al-Khwarizmi Engineering Journal, Vol. 21, No.4, pp. 82- 92 (2025) 86 3.4. Proposed BGA-based cooperative spectrum sensing To enhance detection performance and spectrum utilisation in CRNs, a binary genetic algorithm (BGA) was employed to optimise the weighting coefficients used in the SDF process. The proposed BGA-based model operates by evolving a population of candidate weighting vectors through successive generations, guided by fitness-driven selection, crossover and mutation operations, selection and genetics. The model is capable of addressing straightforward tasks (e.g. image generation and highly complex challenges), including deep learning optimisation and stochastic problem solving. GA deals with a set of possible solutions, each representing a particular answer to the problem in question. The procedure starts with an initial random population of the encoded solutions termed genomes. Next, these solutions are evolved to different generations using biologically motivated operations for selection, crossover and mutation. The quality of each genome is assessed using a fitness evaluation function relevant to the specific problem. If the stopping criterion has not yet been met, individuals with higher fitness values are selected to form the next generation. These individuals undergo crossover to exchange genetic material and mutation to introduce variability. The offspring produced by this process are then evaluated, and the most promising amongst them are carried forward. The abovementioned evolutionary loop continues until a predefined condition is satisfied (e.g. reaching a maximum number of generations or achieving a target performance). A conceptual overview of this procedure is shown in Fig. 4. The strategy generates a seed group of potential solutions and normalises them to meet restrictions, aiming to optimise identification effectiveness by determining the ideal collection of grading vector elements. If the algorithm exceeds the specified limit, the most effective scalar values are then selected. Let C S and L1, L2…LC be the soft decisions of S1, S2…SC, respectively, on the presence of signals, where C is the complete number of SUs. Yi represents the weighting vector of the ith individual that consists of Y1, Y2….YC. In the case of ||Yi ||=1, dp is the detection rate. The fitness value for the ith individual is defined as: . The objective of the selection step is to use crossover and mutation to identify the most advantageous genomes for replication. A higher fitting value indicates a superior option: Funi=dp(Yi). …(12) Here, the Roulette Wheel selection approach is employed. The notation pi represents the probability of choosing the ith individuals. 𝑑𝑖 = FUN𝑖 ∑ FUN𝑖 𝑖𝑛𝑑𝑖𝑣𝑖𝑑𝑢𝑎𝑙𝑠 𝑖−1 …(13) Fig. 4. Flowchart of the proposed GA Based on privilege, the genes that have the greatest likelihood of occurring will be passed down to the following population. Crossover occurs following the selection procedure. The crossover procedure commences by pairing the chosen genes from the existing generation to produce a new progeny population. These partnerships, equivalent to biological parents, are selected randomly out of the current pool via a statistical picking technique. A randomly selected function, such as rand, is then Ahmed A. Thabit Al-Khwarizmi Engineering Journal, Vol. 21, No.4, pp. 82- 92 (2025) 87 employed to ascertain candidate numbers, which are subsequently rounded to the nearest integers to align with real chromosome positions within a certain population size (e.g. 5 individuals). After selecting both parental genomes, a multipoint crossover approach is employed, in which genetic material is exchanged between pairs at several randomly established crossover locations. This process mimics the biological process of genetic recombination, thus promoting diversity and helping the algorithm explore a broader solution space. In the tested configuration, sample random values (i.e. (0.67, 0.23) and (0.12, 0.91)) are generated to guide the selection and crossover points. The resulting selected chromosomes for crossover are indexed as x = [2, 4] and y = [5, 1], representing the parental combinations for that iteration. The complete crossover and mutation procedures are depicted in Figs. 5 and 6, respectively. These figures offer a visual analogy similar to shuffling and splitting strands of DNA to create new genetic expressions in the next generation of solutions. Fig. 5. BGA crossover operation Fig. 6. The operation of mutation In the case of BGA mutation, all the parameters that might be changed are ≈ (mutation rate * population * bit size * number of variables). At a pace of a single chromosomal bit at a time, the suggested bit switches into the matching bipolar digit after the row and column numbers are randomly selected. The process by which BGA mutations work is presented in Fig. 4. The mutation procedure contributes to the creation of a new search space. In summary, the following is how the suggested GA-based system for collaborative spectral monitoring operates from A to G: A: Place t is set to 0 to produce an arbitrary population of genomes. Every population genome is (C*X) bits long, where C represents the overall number of SUs in the system and X represents the bits in every genome. B: Each chromosome will be decoded into an array that corresponds to it in the arbitrary group. The identification accuracy is enhanced via the weighted parameters set of Y= [Y1 Y2 Y3,…Yi]T; Y ≥ 0 satisfying the criterion ||Yi||=1. C: The weighted parameter array should be normalised by combining: D: Most suitable genomes (P*α) will be identified (where α is an amount specifying a percentage of P that is below one) by calculating the value of fitness of each balanced decoded weighted vector (Y) and ranking them based on this value. �⃗⃗� = �̅� √∑ (𝒀𝒊)𝟐 𝑪 𝒊=𝟏 , …(14) E: Return the updated solution (P*(1-α)) via the GA’s selection, crossover and mutation processes and set t=t+1. F: Combine the recently replicated chromosomes with the most promising parental chromosomes to create an additional batch of genes. G: Following the procedures outlined in Steps B and C, decipher and standardise the genetic material from the freshly created population. Repeat step D to determine every chromosome’s value for fitness, then finish the iteration if t equals the preset value ngener. If not, proceed to E. 4. Experimental Scenarios To determine the ideal arrangement for GA, numerous tests were executed by systematically altering essential factors. The assessment concentrated on the impact of these setups on detection accuracy and overall performance. The examined factors included mutation rate, population size, bits per variable and crossover rate. The GA method was executed with multiple settings of an identical factor to determine the ideal array of values. The experiments were conducted using MATLAB®. Ahmed A. Thabit Al-Khwarizmi Engineering Journal, Vol. 21, No.4, pp. 82- 92 (2025) 88 4.1. Parameter tuning and experimental setup The parameters utilised for testing are listed in Table 2. The best-performing configuration, based on detection probability and convergence speed, is shown in Table 3. Table 2, Testing parameters Variable Value Rate of mutation [0.02, 0.12, 0.17, 0.21, 0.28, 0.31, 0.4] Size of population [6, 12, 18, 24, 30] No. bit per variable [2, 4, 8, 16, 32] Rate of crossover [0.52, 0.54, 0.62, 0.64, 0.8] Table 3, Obtained best results Variable Value Rate of mutation 0.12 Size of population 30 No. bit per variable 16 Rate of crossover 0.64 4.2. Simulation results and observations Using the optimised parameters, simulations were conducted over 250 generations with population sizes of 35 and 24 SUs. The mutation probability, crossover portion and false alarm rate (Pf) were set at 0.05, 0.75 and 0.01, respectively. Using the best settings determined in the preceding section, the simulation will run the suggested GA approach. The value of fitness determines its identification likelihood, as shown in Fig. 7. It is observed how the GA option meets the greatest achievable answer, which is 100%. The subsequent settings were employed by BGA as follows: 250 generations to come, 35 population length as minor with 24 SUs, 0.05 chance of mutation, 0.75 population for gestation and 0.01 for Pf. Fig. 7. Sample of obtained results In Fig. 7, data 1 represents the results of the proposed method, while data 2 is the mean fitness binary. The effect of mutation rate on overall performance is clearly noticed in Fig. 8, which shows direct proportionality between the detection rate and the mutation rate. Similarly, direct proportionality between the generation number and the fitness function is shown in Fig. 9. Fig. 8. Direct proportionality between the detection and the mutation rate Fig. 9. Relation between the generation number and the fitness function. 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.8 0.82 0.84 0.86 0.88 0.9 0.92 0.94 0.96 Effect of Mutation Rate on Detection Rate Mutation Rate D e te c ti o n R a te 50 100 150 200 250 300 350 400 450 500 0.8 0.82 0.84 0.86 0.88 0.9 0.92 0.94 0.96 0.98 1 Improved Fitness Progress of GA Method Number of Generations F it n e s s Ahmed A. Thabit Al-Khwarizmi Engineering Journal, Vol. 21, No.4, pp. 82- 92 (2025) 89 Fig. 10. Energy consumption vs. population The relationship between population size and energy consumption in the proposed GA framework is shown in Fig. 10. As the population size increases, the energy consumption exhibits a proportional growth due to the higher number of individuals being processed in each generation. This trend is also linked with the high computational snowball associated with larger populations. Thus, there is a trade-off between improving detection performance and minimising energy costs when implementing real-time detection. 5. Proposal and Other Methods: Comparative Analysis To test the efficiency of the proposed GA-based algorithm, several standard spectrum-sensing techniques were considered for comparison: energy detection (ED), matched filter (MF) and convolutional neural networks (CNNs). The evaluation was done under common simulation parameters of 8 MHz bandwidth, 10 ms sensing time duration and 10 dBm primary transmission power. The system was assumed to be under a very harsh noise environment of SNR ratios below −15 dB. 5.1. Detection rate vs. false alarm rate As shown in Fig. 11, the proposed GA approach achieved a pd of 90% with a pf of 0.04, outperforming CNNs (83%), mf (70%) and ED (58%). This outstanding performance underscores GA’s strength in optimisation and adaptability in the sensing process. Fig. 11. Comparative analysis of the suggested and traditional methodologies 5.2. ROC performance comparison The automated technique provides better detection probability over a wide range of false alarm rates, as seen from the ROC curves in Fig. 12. This is further evidence of the reliability and robustness of the proposed model under varying operational conditions. For the comparison, we used standard deflection factor (SDF), normal deflection coefficient (NDC), modified deflection coefficient criterion (MDC), maximal ratio combining technique (MRC0 and equal gain combining technique (EGC). Fig. 12. Comparative analysis of the proposal against alternative methods 5.3. Bit error rate analysis As shown in Fig. 13, the BER of the GA model was markedly least (10⁻⁵). This result further validates the method’s resilience to noise and channel degradation, particularly in low-SNR conditions. 5 10 15 20 25 30 1 1.5 2 2.5 3 3.5 4 4.5 Population Size E n e rg y C o n s u m p ti o n ( a rb it ra ry u n it s ) 0 0.01 0.04 0.07 0.1 0.14 0.16 0.2 0.4 0.8 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 False Alert D e te c ti o n R a te ED Matched filter CNN Proposed Method 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Probability of False Alarm (Pf) P ro b a b ili ty o f D e te c ti o n ( P d ) Enhanced ROC Performance Comparison of Detection Methods Proposed GA NDC MDC MRC EGC HDF(OR-Rule) Ahmed A. Thabit Al-Khwarizmi Engineering Journal, Vol. 21, No.4, pp. 82- 92 (2025) 90 Fig. 13. BER of the proposed system The proposed autonomy sensing method achieved equally or more competitive results than relative traditional techniques, rendering it quite versatile and adaptable to next-generation CRNs. 6. Conclusion A methodology for spectrum sensing and resource allocation in CRNs through biological algorithms has been proposed in this paper in detail. BGA performance was evaluated for different settings and compared with other conventional spectrum-sensing approaches, such as ED, MF and CNNs. The results of the simulations demonstrated the advantage of the GA-based system in terms of detection accuracy with minimised false alarm rates along with low noise and channel variations. In this research, we also developed a cooperative sensing model based on the GA, which enhances the performance of the soft-combining fusion scheme in the following ways: (1) fitness-based selection, multipoint crossover and (2) adaptive mutation for optimising the weighting vector for secondary users involved; (3) validation of performance through detailed simulation studies, which showcases improved detection reliability even at low signal-to- noise ratios; and (4) traditional versus deep learning methods are compared, and the performance of the GA model is validated against these algorithms. The results demonstrate the ability of GAs to perform dynamic cognitive spectrum optimisation, while also emphasising that fine-tuning genetic parameters, such as population size, mutation rate and crossover frequency, may have drastic consequences for performance. References [1] A. Thabit, "Simulation of a new algorithm to enhance the spectral efficiency of 5G for IoT applications," in Emerging Technology Trends on the Smart Industry and the Internet of Things, pp. 17–26, 2022. [2] M. Alrabeiah and M. S. Alouini, "Deep learning-assisted optimization for spectrum management in cognitive radio networks," IEEE Communications Letters, pp. 1446– 1450, 2020. [3] A. Arshad and S. H. Shah, "A multi-objective genetic algorithm for spectrum management in cognitive radio networks," IEEE Access, vol. 9, pp. 46315–46324, 2021. [4] S. E. Abdelbaset, H. M. Kasem, A. A. Khalaf, A. H. Hussein, and A. A. Kabeel, “Deep Learning-Based Spectrum Sensing for Cognitive Radio Applications,” Sensors, vol. 24, no. 24, Dec. 2024, doi: 10.3390/s24247907. [5] Y. Xu, K. Zheng, X. Liu, Z. Li, and J. Liu, “Cognitive Radio Networks: Technologies, Challenges and Applications,” Feb. 01, 2025, Multidisciplinary Digital Publishing Institute (MDPI). doi: 10.3390/s25041011. [6] A. Alkhayyat, A. Thabit, and A. Adil, "WBAN healthcare-based: Modeling signal to interference ratio with different MAC protocols," in "Proc. 2nd Int. Conf. Engineering Technology and Their Applications (IICET) ", Islamic University, Al- Najaf, Iraq, 2019. [7] A.J. Al-Askery, F. S. Hasan, and A., A., Thabit, " Investigating the Performance of Coded GSIM DCSK Communication Systems over Multipath Rayleigh Fading Channel," Journal of Communications Software and Systems, vol. 20, no 4, pp. 298–306. [8] S. Dhanasekaran and S. Praveena, "Genetic algorithm-based dynamic spectrum access in cognitive radio networks," Journal of Communications and Networks, vol. 23, no. 1, pp. 53–62, 2021. [9] Yun, D.-W.; Lee, W.-C. "Intelligent Dynamic Spectrum Resource Management Based on Sensing Data in Space-Time and Frequency Domain." Sensors, vol. 21, no. 16, 2021, p. 5261. [10] A. Elgendy and A. Ammar, "Spectrum management optimization in 6G wireless networks using genetic algorithms," IEEE Communications Magazine, vol. 59, no. 6, pp. 144–150, 2021. 0 1 2 3 4 5 6 7 8 9 10 10 -6 10 -5 10 -4 10 -3 10 -2 10 -1 10 0 Bit Error Rate for proposed system Eb/No (dB) B E R Proposed system QPSK Ahmed A. Thabit Al-Khwarizmi Engineering Journal, Vol. 21, No.4, pp. 82- 92 (2025) 91 [11] M. Farooq, S. Asghar, and A. Hussain, "Adaptive spectrum management using genetic algorithms in cognitive radio ad hoc networks," IEEE Access, vol. 9, pp. 119834–119849, 2021. [12] A. Gharaibeh and M. Younis, "Genetic algorithm-based power control and spectrum management for cognitive radio networks," IEEE Trans. Cogn. Commun. Netw, vol. 5, no. 3, pp. 426–435, 2019. [13] S. Rama, S. Muthukumaran, V. Sundararajan, and R. Subramanian, "Jellyfish search optimization with deep learning driven autism spectrum disorder classification," Computers, Materials & Continua, vol. 74, no. 1, pp. 2195– 2209, 2023. [14] X. Hu, X. Wang, and X. Zhang, "Genetic algorithm-enhanced spectrum management for 5G cognitive radio networks," Wireless Communications and Mobile Computing, vol. 2020, Art. no. 9782176, pp. 1–10, 2020. [15] L. Huang and Q. Zhao, "Optimal spectrum management in cognitive radio networks using hybrid genetic algorithms," Journal of Communications and Networks, vol. 23, no. 2, pp. 132–144, 2021. [16] A. Jain and R. Gupta, "Performance analysis of genetic algorithm for spectrum management in cognitive radio networks," Journal of Communications Technology and Electronics, vol. 66, no. 6, pp. 550–561, 2021. [17] F. Jiang and P. Wang, "Genetic algorithm-based spectrum optimization in cognitive radio networks: A review," Wireless Personal Communications, vol. 123, no. 4, pp. 3671– 3691, 2022. [18] M. A. Khan and H. Rehman, "A novel genetic algorithm for joint spectrum management in 5G cognitive radio networks," IEEE Access, vol. 7, pp. 67832–67842, 2019. [19] D. R. Kiran and K. Anjaneyulu, "Spectrum management in cognitive radio networks using genetic algorithm and reinforcement learning," IEEE Systems Journal, vol. 15, no. 3, pp. 3948–3957, 2021. [20] A. Kumar and N. Saxena, "Genetic algorithm- based spectrum management for IoT-enabled cognitive radio networks," IEEE Trans. Ind. Informat, vol. 18, no. 3, pp. 1679–1688, 2022. [21] Y. Li and Y. Liu, "A dynamic spectrum management approach based on genetic algorithms in cognitive radio networks," Mobile Networks and Applications, vol. 24, no. 4, pp. 1189–1199, 2019. [22] X. Liu and H. Song, "Spectrum allocation in cognitive radio systems using genetic algorithms and fuzzy logic," IEEE Access, vol. 9, pp. 15137–15148, 2021. [23] S. Nadeem and M. Aslam, "Optimization of spectrum management using genetic algorithm in cognitive radio sensor networks," IEEE Access, vol. 8, pp. 39250–39261, 2020. [24] A. Yadav and S. Rathore, "Genetic algorithm- based spectrum management in IoT-enabled cognitive radio networks," IEEE Access, vol. 10, pp. 4886–4902, 2022. [25] S. Zubair and S. Ahmed, "Joint spectrum and power management in cognitive radio networks using genetic algorithms," IEEE Communications Letters, vol. 27, no. 9, pp. 2107–2111, 2023. ( 2025) 82-92، صفحة 4، العدد21مجلة الخوارزمي الهندسية المجلد احمد عبد الرضا ثابت 92 األداء وتحسين إدارة الطيف في أنظمة االتصاالت باستخدام الخوارزمية الجينية تقييم 4حيدر وهب عليوي، 3علي جابر العسكري ،2بسام حميد عبد ،*1احمد عبد الرضا ثابت ، بغداد، العراق الجامعة التقنية الوسطى الكهربائية، الكلية التقنية الهندسية 1،3 ، بغداد، العراق الجامعة التكنولوجية الكهربائية، قسم الهندسة 2 قسم الهندسة اإللكترونية والكهربائية، جامعة برونيل لندن، المملكة المتحدة 4 ahmed.thabit@mtu.edu.iqالبريد االلكتروني: * المستخلص زدحام ونقص في ظل تزايد تقنيات االتصاالت الالسلكية، مما يزيد من الطلب على طيف الترددات الراديوية المحدود، ويخفف حتى اآلن من حدة مشاكل اال ، القادرة على تحسين دقة الكشف والوصول (BGA) الطيف يعتمد على الخوارزمية الجينية الثنائيةاالستخدام والتداخل، تقدم هذه األطروحة نموذًجا حديثًا إلدارة CRN الترميز الثنائي لتحديد عوامل الترجيح المثلى للمستخدمين الثانويين في سيناريو BGA تتبع .(CRNs) التكيفي للطيف في شبكات الراديو المعرفية دمين الثانويين نتائج بأداء أسرع بكثير وموثوقية أفضل من األساليب الجينية التقليدية. في مخطط استشعار الطيف التعاوني المقترح، سيُرسل العديد من المستخ على ضبط معامالت الترجيح بدقة عبر آلية دمج القرار الناعم. تتطور الخوارزمية BGA االستشعار المحلية الخاصة بهم إلى مركز دمج حيث سيعمل تحسين ينية إلى آخر من خالل تطبيق عمليات االختيار والتقاطع والطفرة الكتشاف أفضل تكوين. أُجريت تجارب محاكاة مكثفة لدراسة آثار المعلمات الج من جيل الُمحسَّن يُمكنه تحقيق احتمالية كشف تصل إلى BGA الحرجة الحتمالية الطفرة ومعدل التقاطع وحجم السكان على قدرة الكشف. أظهرت النتائج أن إطار عمل يبلغ 96% إنذارات خاطئة يبلغ 0.1تقريبًا، ومعدل يبلغ حوالي 0.12، ومعدل طفرة بتات إلى ⁵⁻ 10× 7، ومعدل خطأ اإلشارة نسبة تكون حتى عندما ة عند اختبارها مقابل خوارزميات قًا واضًحا للخوارزمية المقترحديسيبل. باإلضافة إلى ذلك، أظهر التقييم المقارن تفو 15-الضوضاء منخفضة للغاية عند ء. كما يؤكد تقليدية مثل كشف الطاقة، والترشيح المتطابق، وااللتواءات القائمة على الشبكات العصبية، وذلك عند تعريضها لظروف صعبة ومعرضة للضوضا ، ويقدم حالً قابالً للتطوير إلدارة الطيف في أنظمة الجيل الخامس CR هذا العمل على إمكانية تطبيق الخوارزميات التطورية في تعزيز الذكاء المعرفي ألجهزة .الحالية وأطر الجيل السادس المستقبلية mailto:ahmed.thabit@mtu.edu.iq