Genetic Algorithms-Based Model for Multi-Project Human Resource Allocation Abstract Jing Ai Shijiazhuang University of Applied Technology, Shijiazhuang 050081, China With the constant development of computer and network technology, modern enterprises tend to recruit interdisciplinary talents, who could support the parallel projects operation. This condition poses new demand for human resource allocation (HRA) in an enterprise, as the old-fashioned HRA that allocated one person on one post would fail to maximize the personnel utilization. Therefore, this paper concludes the approaches of integrating multi-project human resources based on genetic algorithms (GA) and then designs a model for the HRA. It aims to improve the human resource management (HRM) to optimize the HRA through applying the GA-based management model for the multi-project HRA. Keywords: GA, Cluster Analysis, Human Resource, Allocation Model. 1. RESEARCH BACKGROUND 1.1 Literature review With modern enterprises transforming from produce-oriented to service-oriented and single sales mode to onand off-line multi-channel operation, the HRM faces new opportunities as well as challenges (Zhong and Shi, 2015). On the one hand, the HRA is the basis of enough human labor in an enterprise, and the HRM is required to integrate multiple projects. As human labor increases, so is the management scope, dramatically enhancing the demand for more detailed management rules (Xu and Liu, 2016). On the other, human resource integration in multiple projects itself may encounter less favorable conditions, so if the technical staff from the single project cannot adapt to the new situation, the performance of the HRM will also be compromised (Yuan and Zhang, 2016). Thus, approaches of the HRA is especially crucial in further researching integration of multiple projects. Project resources explored by enterprises accord with the law of genetic reproduction, hence, if human resources are well integrated the same way genes are for the next generation, the optimal HRM in multiple projects could be achieved, stimulating the enterprise to integrate existing resources by collaborative labor allocation. 1.2 Research purposes Old-fashioned plans or laws of human allocation usually see subjective mistake or error. However, the GA-based method provides alternatives for allocating human resources to natural selection (Lin et al., 2016). In the genetic evolution, the optimizing is attained, with the new generation equipped with all the strength of former generations and their weaknesses removed. Repeated evolution brings about a solution set, among which one is closest to the optimum result, and this is the most desirable result of the HRA (Wu et al., 2015). Therefore, the GA could enable a comprehensive review of the multi-project HRA optimizing, to provide a dynamic mechanism for enterprises in the HRA and improve their decision-making criteria and quantitative indicators. Utilizing the strength of the GA, the paper creates the GA-based model for the multi-project HRA to maximize the efficiency of human resources and integration result. 2. CREATE THE GA-BASED MODEL FOR THE HRA IN MULTIPLE PROJECTS The HRA in multiple projects requires the optimizing of internal structures to achieve the best management, specifically the coordination among project department, resource department and management department. Management office, as the core of management, needs to evaluate the enterprises circumstance and analyze the critical part of the HRM and its allocation while considering departments resources (Shen and Zhao, 2014). The 70
GA also seeks to form an effective mechanism for employees in multi-role to switch successfully. Based on this, the GA s quantitative analysis could optimize the HRA as a whole to improve the work efficiency. 2.1 Problems In multi-project operation, all tasks progress simultaneously, which may cause time conflicts for personnel. The HRA could be the solution to reduce the conflict and complete all the tasks simultaneously. Suppose the employee base is Q=[Q 1, Q 2,, Q m] and each employee undertakes more than one critical responsibility with adequate skills and capabilities, but the level of skills varies (Huang et al., 2014). If multiple projects are classified as A=[A 1, A 2,, A n], and the processing set is B=[B 1, B 2,, B k] in the n th procedure, then the processes are determined by overall workload and skill type, i.e. project duration (d)=project time (A*worker number). The allocation of staff skill sets in multiple projects are shown in Figure 1. Skill X1 Coefficient Y1 Skill X f Coefficient Y f Staff member i Skill X2 Coefficient Y2 Skill X3 Coefficient Y3 Figure 1. Allocation of Staff Skill Sets All the processes in this project can be considered work completed within working location, while each work is undertaken by staff with different skills. When staff Q 1 is equipped with i skills with the skill set [i=1, 2,, m], the competence coefficient of individual employees may vary. This paper uses y 1, y 2,, y f as competence coefficient, in which f represents the current types of skills staff possess (Shen and Wang, 2013). If 1 represents high level of skill competence, and 0 low level, then the coefficient of the rest skill types can be seen as left unapplied and could be transferred to other projects. Meanwhile, if more employees are assigned to assist this certain employee but the project remained unfinished, then he could be transferred to other projects according to the GA. 2.2 Model of the HRA in multiple projects Through the multi-project HRA, the enterprise aims to realize the simultaneous operation of several projects and finish in time. If project operation sees time redundancy, it means the project duration is dramatically reduced or delayed (Li and Lu, 2013). Therefore, the HRA model is supposed to satisfy the condition for priority selection and dominated by decision-making variable for optimal allocation, ensuring that all projects can produce a certain quantity at a certain speed. If the resources optimizing level reach or exceed the expected state, the objective could be achieved in advance (Wu et al., 2013). This situation cannot be accomplished overnight as it is subject to many factors in operation. As a result, whether the allocation model could achieve the goal depends on the level of the optimal solution, to eventually realize a balanced development for the multiple projects. Suppose an enterprise is performing the same procedures in multi-project operation, key techniques are represented by A, B and C, project processes by 1, 2 and 3, and process code by i and j, then proper conditions can be found in i=1, 2, 3,, n and j=1, 2, 3,, k. The objective function is: n n k j=1 max F(m) = max[f(x) + f(t)] = i=1 + f(t) (1) There are three conditions for this objective function. First, order constraints apply to tasks within one project but not between projects. Each procedure requires at least three employees to cooperate to complete a single n i Z ij 71
project. Second, the project release and completion time have been set, and no processes can be halted in multi-project operation (Zhang and Zeng, 2013). Third, each procedure requires only one skill to accomplish. If a procedure has to be halted for lack of suitable worker, replacements can be transferred from other tasks to this procedure. Let max[f(x) + f(t)] be the sum of the average skill coefficient of staff, the larger the sum is, the higher quality the project is likely to achieve. In a single project, n represents the number of staff allocated, suitable skills are identified in the j th employee from Z ij. Suppose m is the i th key point in completing the project and multi-task value indicates the speed of the project B i=m i-a, then the dispatch and sequencing of the project should be considered from the beginning and the difference within some time is expressed as: f(t)=min(b i)-t i, in which T i represents the completion time, and the level of completion is seen as the reflection of time utilization. 3. ANALYSIS OF THE GA-BASED FOR THE MULTI-PROJECT HRA As shown in Figure 2, the GA consists of coding, initial population calculation, genetic operation, variation, and termination. 3.1 Coding Figure 2. Genetic Algorithm Flow of Multi-project Human Resource Allocation Coding selects real parameters in the GA. Better coding plans are available in a complete operation mechanism and can be extended to multiple projects so that the strength of the HRA can be observed. It involves floating-point number coding, binary coding, character coding, etc. (Luo and Fu, 2014). The binary encoding adopts computer binaries to connect the gene properties and seeks a solution to the problem from the string. Displaying the minimum characters, it facilitates calculation in crossing and variation and can be applied to decoding. Grey Code which is also seen as the extension and transformed representation of binary coding could 72
complement binary coding in its structural weakness to make it more complete and consistent. It can improve local search capacity and complete concurrent genetic operations like crossing and variation, in addition to some minimum character set. 3.2 Initial population Initial population calculation randomly selects from the initial population to produce optimization. It excels in calculating the individual s fitness and retains the next generation with some corresponding adjustments. Those with higher fitness will be more likely to be reserved, while gene group s lower fitness reveals a removal possibility. 3.3 Genetic operation The genetic operation is the realization of the GA through selection, cross, and variation. The process meets fitness selection and manages to produce the next fittest generation as the optimal solution for an energetic population (Hou et al., 2011). The optimal solution set is the most practical mode in integrating human resources. Using the GA to get the minimum probability can verify the most suitable or similar theoretical value from the most optimal solution in the set. 3.4 Gene cross In this procedure, new-born genes interact with others, which is gene-gene interaction to produce new genes in a basic selection. If two near standard parameter sets are obtained from the higher probability, gene traits could pass through cross calculation from parent gene to the child gene. The matching mode also follows the optimization of N individuals, i.e., N/2 range. 3.5 Variation Variation is the basis of a new generation. Variation calculation which adopts the floating-point or binary coding helps maintain the diversity of an individual. The probability value for basic variation condition is taken as the threshold value and it randomly selects conditions for variation calculation from the locus (Zhang et al., 2010). Then, uniform variation method could be applied to achieve random data selection within a range of even distribution, so that a fairly low probability value could be used to replace an original gene value. Termination standards are to be established with the GA. If the conditions for the most optimal solution cannot be met, the process returns to the initial stage to repeated calculation, until reaching an optimal solution for gene combination. Then the GA could be terminated. But any result is a theoretical value that is infinitely near the most optimal solution, so the maximum evolution number is generally adopted for calculation. 3.6 Evolution operation Evolution is an extension of an optimal solution, whose calculation involves selection, cross, and variation. Its selection could be operated with roulette, in which individuals are selected from the parent group to be hybridized, accounting for 1/2 of the parent generation. Let f i be the fitness level of chromosome i, its calculation model is: The probability for chromosome to be selected is: f sum = n k=1 f k (2) A i = f i f sum (3) In the cross operation, the one-point crossover can also produce two new genes of the next generation and gain the maximum value in cross probability (Gao and Chen, 2010). If the threshold value range is small, new gene generation will slow down; but if it is too fast, group superiority will be compromised. Therefore, the range of (0.6-1.0) is appropriate. In the variation, however, it is mandatory to classify the chromosome of the next 73
generation using pre-set probability whose range is (0.01-0.1). Suppose the length of the chromosome is M, with the single gene variation probability of 1/M, then an allocation mode that is nearest to the most optimal solution could be obtained by defining evaluation standards, providing references for multi-project personnel transfer. 4. CONCLUSION In conclusion, multi-project operation requires the HRA and using the GA, a natural selection mechanism, to come up with the most optimal personnel allocation could maximize the personnel utilization. Applying the GA-based model, one also has to consider the circumstance of individual enterprises. The GA is calculating and raising a plan that is infinitely closer to the most optimal solution set, but not the absolute optimal one. Therefore, the GA-based model for multi-project HRA is only for theoretical reference. The real-life personnel allocation needs to fully consider the specific requirements before reaching the most optimal solution for the allocation. The most critical standard is whether staff could meet the skill demands of other posts. If they could, personnel allocation will be easier. Thus, putting this model in practice requires constant stage training for staff, focusing on capacities like computer technology, network operation, information technology, etc. and developing interdisciplinary talents to support enterprise to expectedly complete projects. REFERENCES Gao S., Chen Y.J. (2010). The genetic algorithm of cloud adaptive genetic algorithm, Logistics technology, 29 (20), 95-97. Huang X.M., Luo S.G., Wang Y. (2014). Hr support efficiency and structure of strategic emerging industry development - taking Guangzhou as an example, Journal of Jingchu institute of technology, 29 (02), 76-81. Hou L.L., Yao Y.N., Fan S.D., Jiang P. (2011). Research on human resource optimization of ship overhaul, Journal of Wuhan university of technology (transport science and engineering), 35 (02), 325-328. Li Y.S., Lu W.X. (2013). Application of genetic algorithm for viral evolution in locomotive maintenance system, Industrial control computer, 26 (09), 106-107+109. Lin J., Zhou G.H., Xia F.L., Wang F., Wu C.X. (2016). Research on the influence of personnel flexibility on scheduling of job shop, Computer application research, 33 (10), 3017-3020+3025. Luo X., Fu X.W. (2014). An improved genetic algorithm applied in performance appraisal, Technology communication, 4 (20), 189+175. Shen G.J., Wang J.Y. (2013). Construction of human resource allocation model based on multi-objective hybrid genetic algorithm, Statistics and decision, (21), 60-63. Shen G.J., Zhao X.L. (2014). A human resource assignment model and method based on improved genetic algorithm, Statistical and decision-making, (16), 52-55. Wu H., Yang J., Wang H.Y., Yin D.M., Cui B., Wang Y.X. (2013). The multi-objective and acoustic search algorithm for solving human resource allocation problems, Computer technology and development, 23 (02), 65-68+72. Wu M.T., Liu Z.Y., Jiang L.X., Cai J.L., Yan S.X. (2015). The application of BP-GA algorithm in cost control - based on JNC co., LTD, Technology innovation and application, (24), 28-29. Xu M., Liu S.W. (2016). Based on genetic algorithm and guidelines, the optimization design of high-rise building structures, Architecture knowledge, 7 (07), 125-127. Yuan F.J., Zhang J.P. (2016). Human resource management based on multi-stage genetic algorithm, Journal of Yunnan national university (natural science edition), 25 (03), 275-279. Zhang W., Zhao J., Chi H. (2010). The human resource portfolio evaluation model based on genetic algorithm, Journal of Tianjin university of technology, 29 (06), 85-88. Zhang Z.L., Zeng C. (2013). Research on human resource scheduling of software project based on improved genetic algorithm, China new communication, 15 (02), 87. Zhong F., Shi S.Y. (2015). Research on human resource supply chain platform based on hierarchical genetic algorithm, Electronic design engineering, 25 (05), 1-4. 74