Abstract

This paper addresses the design of an enterprise human resource management system due to multivariate statistical analysis computation in a random matrix recommendation algorithm in a distributed scenario. This paper defines multivariate statistical analysis human resource practice (DI-RP). It determines the composition of DI-HRP based on the nature of multivariate statistical analysis and enterprise human resource practice. In addition, the role of DI-HRP in influencing employees’ innovative behaviors is explored based on the theoretical basis of resource conservation theory. This paper mainly develops according to the software development process typical to software engineering, organizes the current business logic of the company, understands the relevant content of the company’s human resource management, conducts requirement research on the human resource department, analyzes the feasibility of system implementation from different perspectives, and finally designs a human resource management system based on B/S architecture on the result of requirement analysis. The technology and tools used in the system were decided on the existing technical architecture of the company. The system design was divided into five modules: personal information management module, work management module, attendance management module, reimbursement management module, and entry/exit management module according to the requirements and finishing process, and each functional module of the system was coded and implemented, respectively. The development tool is PyCharm, and some front-end pages are edited and modified by Visual Studio Code. The Permission model of Django is used to add corresponding permissions for various types of users to ensure the security of the system when running.

1. Introduction

In the current social context of the times, enterprises should put forward more convenient and faster requirements for human resources management with the development of the information age. Still, most enterprises need to strengthen and improve the management system of human resources management system, optimize the allocation of human resources, carry out specific unified control to centralize management, and break the limitations of the previous human resources management. The end-users of the HR management system can access it through computers and mobile terminals so that the information interaction of human resources can become timely and smooth, which can realize the improvement of the efficiency of the enterprise and effectively realize the established planning policy of the enterprise, which can promote the development of the enterprise in the expected direction, avoid the waste of resources of the enterprise, and enable the employees to play their initiative fully [1]. The application of information technology in human resource management is mainly used as a tool; as the saying goes, if you want to do a good job, you must first make use of the tools, and if you are going to do an excellent job in human resource management, you can achieve twice the result with half the effort by using a convenient and efficient human resource management system. The application of information technology in HR work can significantly improve the efficiency of the HR department [2]. It will substantially reduce the proportion of routine tasks that take up the time of HR managers. The above are the primary objectives of many enterprises to build HRMS, followed by the need to optimize business processes and improve service quality.

Distributed network environment has excellent performance, such as openness, flexibility, and scalability, which helps to reduce cost and improve business agility, and is suitable for many types of application deployment. Distributed networks are rapidly developing under new technologies such as cloud computing and blockchain. They have been widely extended to many fields such as digital finance, the Internet of Things, intelligent manufacturing, and supply chain management, actively promoting various industry changes and technological innovation development, and have become a new trend in network development [3]. The terminal data of users is constantly gathered to the cloud server with sufficient arithmetic power and storage and other resources. The cloud server provides data analysis, processing, storage, and management services. In contrast, blockchain provides data traceability and a trust mechanism for the system to solve trust problems in the cloud. However, with the development of the digital economy era, user data faces more severe and complex data security and privacy issues in the complex distributed environment based on cloud computing, blockchain, and other technologies. At present, massive amounts of data are repeatedly subject to data theft and leakage, and many users are worried about the security and privacy of their data. Once a malicious person obtains private data involving users’ personal information, it will cause severe threats to users’ life and property security and harm their minds. Therefore, data security and privacy issues in distributed environments have become the focus of attention in academia and industry. There are still many pressing issues and challenges to be studied.

The data stored in the HR management system are of different types and various structures. These diversified data express multiple information employee in the enterprise, such as age, gender, education, and other essential information [4]. This data information can help enterprise managers to consider various influencing factors when making decisions. At the same time, the traditional HR management system can only store and display data but not scientifically analyze multiple kinds of data. With the help of data mining technology, using algorithms such as regression, clustering, classification, and association rules, we can conduct a comprehensive analysis of HR data and make full use of the diverse characteristics of the company’s employees to achieve objective and comprehensive analysis and assist business managers in making sweeping decisions [5]. The rapid development of computational science makes the processing and analyzing of high-dimensional massive data possible, and theoretical and applied research on high-dimensional information is emerging. Studying the spectral behavior of high-dimensional random matrices supports statistical theory for processing and analyzing high-dimensional data, making high-dimensional data more standardized and precise. The dimensionality of high-dimensional data is sometimes even higher than the number of samples. The programming language plays a pivotal role in the analysis and design of the HRM platform, and the communication between humans and machines depends entirely on the programming language to accomplish this complex action. Some classical multivariate instruments in statistics assume that the dimensionality is fixed, which leads to the relevant conclusions in classical occasions that are not applicable, such as hypothesis testing and principal component analysis, which are standard techniques in classical statistical inference, also facing the challenge of high dimensionality. Extensive data processing methods have emerged based on the theory of high-dimensional random matrix spectra.

Using Internet technology, it is possible to achieve rapid iteration of HRMS while using cloud architecture and big data technology to improve the performance of traditional HRMS and reduce the cost of using the system. Rahman et al. study the promotion of HRMS to enterprises based on the information technology perspective [6]. And based on HRM theory, the design is optimized from the standpoint of HR system management and resource management, combined with HR implementation and systematic construction to improve the efficiency of HRM and by using the system to guarantee the implementation of relevant systems and measures. Li et al. believe that constructing an HR management system helps enterprises implement HR strategies [7]. The system provides a quantitative data basis for the actions of HR strategy. Based on this information, we can quantitatively evaluate the performance of employees, their competency, and the degree of fit and propose training and development plans for employees based on their data. Yang et al. integrate the multilevel gray-scale decision-making algorithm into the employee evaluation module of the HRMS [8]. Kushwah and Ali designed the mechanism for managing, storing, and implementing the retrieval function of employee files and improved the performance of employee file management by adopting an optimized indexing method [9]. Ramachandra and Setturu believe that the core of the HR management system is the processing of employee data, so data capability is the core function of the HR management system [10]. The comprehensive dramatization of human resources is the future development direction of the HR central system. Through play, it provides a decision basis for the company to improve HR management and promote the improvement of HR management level. Hsu et al. believe that the core of the HR management system is performance management [11]. Performance management is the basis of staff management and the cornerstone of motivating staff development. Since companies have different requirements for human resources as the market changes and development, the management system needs to be adjusted accordingly, and it is scalable in terms of assessment methods. Emami et al. studied the authority system of the human resource management system [12]. Since there is an organizational hierarchy in human resource management, different authority management functions need to be implemented.

The HR management system in the enterprise has taken shape, which contains certain HR management information and is also designed for the storage of data. The HR management system has been completed, and almost all HR-related data has been integrated. More efficient data analysis tools and report generation tools were designed, and the goal of information sharing was achieved. The completeness of the HR management system was already relatively high at that time, and the corresponding concept of HR management was mature. Management systems could be developed to match the related business processes for different industries. With the advent of a knowledge-based economy, the view of human capital has been formed with people as the carrier of knowledge [13]. The importance of human capital is beginning to surpass that of capital such as equipment and land. Companies started to choose human resource management systems to maximize their effectiveness in managing human resources more appropriately, so they were already widely used by companies in Western developed countries. To make HRMS development less difficult, other enterprise systems have also reserved interfaces for HRMS design in action, and the HRMS has been vigorously promoted based on Internet technology.

The data are usually assumed to obey a multivariate normal distribution in multivariate statistical analysis. The maximum likelihood mean estimator, i.e., the sample mean vector, is the best estimator for the data. However, Stein et al. found that the total likelihood estimate of the mean vector is not reasonable when the data dimensionality is equal to or greater than 3. Scholz et al. proposed an improved mean vector estimation method for multivariate customarily distributed random variables with a unit covariance matrix, where the data dimensionality satisfies p2 [14]. Later, Bai extended this estimation to the case where the covariance matrix is diagonal [15].

Further work was done by Snihur et al. under the condition that the covariance matrix is unknown, but the data dimension is always smaller than the sample size n [16]. The analysis of high-dimensional data has become a research hotspot, and the estimation of its mean vector has received increasing attention. Under constant quadratic loss, Pal B et al. further investigated their method based on the literature [17]. They proposed a high-dimensional James-Stein type mean estimation method using unbiased estimation of risk differences. Canal D et al. suggested an optimal shrinkage mean estimation method by minimizing the quadratic expected loss function when the covariance matrix is unknown [18]. Ashton et al. derived an optimal shrinkage estimation of a high-dimensional mean vector using random matrix theory [19]. The problem of e estimating the mean vector of a multivariate normal distribution with an unknown singular covariance matrix was also discussed by Yang et al. Various shrinkage mean estimation methods were obtained [20].

3. Distributed Environment Model Analysis

There are many complex issues such as consistency and reliability in a distributed environment, which can be effectively solved by mature technology tools, reducing development costs, and ensuring system quality. This chapter introduces the core tools and technologies used in the project design and implementation process, introducing the features, implementation principles, and application scenarios of Etcd and Redis in turn, to ensure strong data consistency, elegantly handle election problems in distributed environments, and tolerate machine failures to avoid single points of failure [21]. As a distributed component, it provides service discovery, message subscription notification, and distributed locking capabilities in addition to storage functions. It is open-source software written in the Go language, which has excellent cross-platform support and is backed by a robust open-source community. Its main features are as follows: (1) simplicity: read and write data using the standard HTTP protocol, which can be easily accessed by various applications; (2) security: SSL client authentication which can be used; (3) efficient: single instance write speed of more than 1000/s; (4) trusted: through the Raft algorithm to ensure strong data consistency. The features of Redis are a fast, K_V-based data structure. Highly useable and distributed storage emphasizes communication and synchronization between nodes to ensure consistency of data and transactions on each node. Redis is more like an in-memory cache. Although it also has a cluster for master-slave synchronization and read-write separation, consistency between nodes mainly emphasizes data and does not care about transactions, so read and write capabilities are robust, and qps can even reach 100,000+.

The algorithms that play a driving role in the distributed optimization of multi-intelligent systems include the basic distributed subgradient descent algorithm, the fast converging distributed alternating direction multiplier method (ADDM), and the subgradient-push method for directed networks and the short row stochastic distributed optimization algorithm (FROST). Before presenting the relevant algorithms, a model of the relevant distributed optimization problem needs to be constructed. The goal of distributed optimization for a multi-intelligent system is to have all the intelligence cooperates to solve the minimum of the sum of local objective functions. The model of the unconstrained distributed optimization problem is as follows:where denotes the global decision variable, is the local objective function of intelligence , and the provincial objective function is protected so that only the intelligence knows it. The optimal solution to the problem may not be unique, but all intelligence needs to find the exact optimal solution. The unconstrained optimization problem simulates the optimization decision problem in an ideal state. At the same time, the existing system has complexity and uncertainty and is easily restricted by the external environment, so the optimization problem with constraints is widely studied:

As the optimization problem is continuously researched and realistic systems are considered, the constraints become more and more complex, with different rules for each intelligence, and may be subject to multiple conditions. Various distributed optimization algorithms have been proposed in the existing literature to solve optimization decision problems, and the following are some basic classical distributed optimization algorithms.

3.1. Distributed Subgradient Descent Algorithm

Each intelligence continuously updates its state values along a gradient descent by fusing its estimates and those of its neighbors through a weighted average of the weight matrix. The algorithm’s convergence rate is O (log k/k) for general functions and O (log k/k) for strongly convex functions. Many researchers and scholars have designed many deformation algorithms based on the distributed subgradient descent method to solve distributed optimization problems effectively; for example, the distributed subgradient projection algorithm is designed to solve the optimization problem with constraints. The distributed subgradient descent method has a simple form and broad applicability but has the disadvantage of a slow convergence rate.

3.2. Distributed Alternating Direction Multiplier Method

This method is a fully distributed optimization algorithm in which each intelligence updates its decision variables in turn and passes the updated variables to the network. Unlike the traditional distributed subgradient descent method, this algorithm mainly finds saddle points by solving the distributed optimization problem. More importantly, the distributed alternating direction multiplier method enables faster convergence.

3.3. Fast Row Stochastic Distributed Optimization Algorithm

where is the heterogeneous step size, is the random row matrix, and denotes the element of the variable. The algorithm uses random row matrices, where the intelligence can autonomously decide the weights from the neighbor information, and uses gradient tracking techniques and row random matrices to achieve fast and accurate convergence of the algorithm.

4. Random Matrix Model Construction

The theory of random matrices has opened a vast field of research for solving the analysis of extensive dimensional data. The study of random matrices originated from the need for physics development, and it was initially used to describe the energy distribution of many disordered moving particles. Since the empirical spectral distribution function of large-dimensional random matrices has many excellent properties, random matrix theory is widely used to distinguish nonnoisy information from disordered noisy details in a system and thus to identify the unique properties present in the system.

To ensure user data security, users need to generate a random matrix for generating scrambled gradient data before transmitting the gradients. Performance testing is to test the system’s performance parameters and analyze the system’s performance parameters under a different number of requests by simulating the login and use of the system and test whether the system can meet the user’s requirements in real business scenarios. We can find out the hidden problems in the design through performance testing and eliminate them. Therefore, to obtain the complete aggregation result after aggregation of the user’s component gradient data, it is necessary to generate a random matrix that can be aggregated and canceled among multiple users. The basic idea of designing this random matrix is to perturb each user’s input. The current research work is based on a centralized matrix decomposition recommendation algorithm that generates random noise and interferes with the objective function [22]. Considering that there will not be real-time end-to-end communication between users and considering the actual scenario, such as real-time communication between users and users and the number of projects and users, determines the dimensionality of the random user matrix; when the data of participating users and projects are too large, due to the large dimensionality of the random matrix, even if the transmission between users and users will significantly increase the communication spending, to ensure data security, we do not take a trusted third party, but using a pseudo-random function generator between users to generate a high-dimensional matrix is very efficient. To reduce the communication overhead, we do not consider interuser assistance in key generation, and we help users create local random matrices using a recommendation server.

When all users utilize the number generator to generate random matrix pairs among users , from the gradient descent formula of matrix decomposition, the gradient is represented by a matrix with k rows and j columns, j representing the number of items. For the privacy problem about the distributed matrix decomposition model in the previous section, for the original gradient, if the privacy protections of the rise are to be performed, the same k row, j column scrambled matrix needs to be generated. The users can mark and filter sensitive and nonsensitive items according to their needs.

Suppose there is a total item set m, a sensitive item set s, and a nonsensitive item set n; any two users a, b, customize the sharp item set and the nonsensitive item set , of their respective item sets. Take the sensitive item sets of the two users and use them as the dimensions of the scrambling matrix:

Either encryption-based privacy-preserving methods or methods by perturbing data will either lose prediction accuracy or require sacrificing computational performance at the cost of matrix decomposition recommendation algorithms. In this paper, a new distributed matrix decomposition privacy-preserving method is proposed to address the privacy issues in the matrix decomposition of the challenges of distributed recommender systems. The improved matrix decomposition recommendation model is mainly from user privacy. The main design idea is to ensure that the user’s historical ratings are always saved locally by distributed model training. The user also generates the recommendation results locally, providing the user’s data security to the maximum extent [23]. At the same time, some parameters may lead to privacy leakage during the algorithm execution; we design a new noise addition algorithm for privacy protection, as the whole system model for training considers the frequent communication between the user terminal and the recommendation server, with each iteration to generate the noise matrix generated by the amount of computation, combined with dimensionality reduction technology to compress the data transmission. The random matrix decomposition process is shown in Figure 1.

Each user will train the matrix decomposition model on their respective device and complete the recommendation by iteratively computing the gradient aggregation update. As seen in the matrix decomposition model in 3.2, iterative feedback between the user and the server is required in the distributed gradient descent algorithm. The computation relies on the individual user’s ratings . This means that individual users’ gradient updates can be computed locally without uploading their user gradients or rating data to the server. Before the model training starts, the recommendation server needs to initialize the matrix V and send V to all users who initialize the U matrix locally. Local users use matrix V; the locally saved scoring data is updated against :

5. Multivariate Statistical Analysis of Enterprise Human Resource Management System Design

The design and development of HRMS are based on B/S architecture. The system uses PyCharm and related components as development tools for functional development; my SQL database management system is used for the system’s background data storage, and the system development platform is Mac OSX. The system is architecturally divided into three layers: representation, business logic, and data. The system architecture is shown in Figure 2.

The system uses B/S architecture; users need to connect to the system through the network and keep connected through Internet exchange routing; the primary binding protocol involved is HTTP protocol; the server-side mainly consists of two parts, which are Web server and database server, where the Web server mainly deals with the corresponding business process. The Web server can directly access the database server. The internal nodes of the company include all the computers connected to the intranet. Users can access the system through any of these computers. Since, in some exceptional cases, access mechanisms are provided for external users who are not connected to the company’s intranet, users can access through VPN communication. They must pass through the firewall set by the company to ensure the security of data information.

LAMP is the initials of Linux, Apache, My SQL, and PHP/Perl/Python; the LAMP system has good performance and stability, and it is popular. Regarding the Script language of the Web server, PHP, Perl, and Python all have their supporters. LAMP is a group of servers usually used together to run dynamic websites or servers. The number of projects developed with PHP is quite large, and there are some notable and joint large-scale projects, such as Word Press (Blog), Joomla (Content Management System, a branch of Mambo), PHP BB (Bulletin Board), and Media Wiki (Wiki developed and used by Wikipedia). Based on the premise of free and open access, this study will also use LAMP as the backbone of system development [24]. In the analysis and design of the HRM platform, programming language plays a pivotal role, and the communication between humans and machines depends entirely on the programming language to complete this complex action. The positioning of the programming language in a system is like the positioning of blood in the human body, without which the system cannot function at all. The smooth flow of blood in the human body will significantly impact a person’s health; similarly, whether a programming language can fully perform its function will also affect the information system. The topology of the system network is shown in Figure 3.

In terms of programming language, this study does not just write code in the original PHP but uses the Cake PHP development framework to build the prototype of this HRMS platform; taking advantage of its inherent MVC modular architecture, the system is forced to form itself with a complete modular application architecture at the beginning of the design development. A system object like this is its object-oriented concept, and its benefits are brought to the fore in the MVC architecture. With a strong development team and community behind it, you can enjoy the world’s most advanced Web development technology with a single network line. The infrastructure will be upgraded and revised over time. Suppose you encounter a problem that cannot be solved. To ensure user data security, users need to generate a random matrix for generating scrambled gradient data before transmitting the gradients. Therefore, it is necessary to generate random matrices that can be aggregated and offset among multiple users to obtain the complete aggregation result after the aggregation of the user’s component gradient data. In that case, you can still seek the assistance of the development team or community on the Internet so that the system created by the institute will not become a canoe, which will be of great help to the maintenance of the system in the future. Based on the concept of object-oriented system design, we do not need to take the trouble to design a set of object-oriented systems from the beginning to the end; we follow the rules of this framework to build a solid object-oriented system. Before starting to design this HRMS, the author of this paper will do a complete comparative analysis between the traditional system development model and the MVC system development model used in this study. The specific PHP code is all mixed and concentrated in one entity’s code; in other words, when it is stored on the hard disk, it is just a Script file; the architecture seems to be very simple, but in fact, it will make the future maintenance and error removal much more difficult. MVC design pattern can facilitate the division of labor among developers, improve development efficiency, enhance the maintainability and expandability of the program, and use the controller to separate the model and view, reducing the coupling between them. This file covers all functions, including the following:(1)Receive instructions from a URL or a link from the client-side and calculate them(2)Send the operation result directly to the database and request the database (Request)

The programmer can even use “Include” to include other people’s files in the code he designed for convenience to save the effort of writing the program, which makes the chance of program error higher again. In one of the most complex PHP codes, the variables in the code can be called anywhere and modify its value and setting; however, it is very accessible when writing the program; you can change it as you like, but once the program is wrong and you want to get rid of the error, you must spend more effort to find out the error, which is very unfavorable in the development of the system. This is very unfavorable in the development of the system, and the system developer should not sacrifice the stability because of saving effort. Another disadvantage is that the same code may be written repeatedly in different places, which is a significant concern of traditional writing. This study does not use this relatively primitive way to build the system. The available statistics of the HRMS usage are shown in Figure 4.

The front-end display system concentrates on processing user requests, displaying system functions, and the front-end display window of user rights information. It can provide information on the processed data returned by the application layer. The development of the portal system does not involve the development of logic between functional modules, so it is a system with minor development difficulty and workload, and the principle that developers treat the portal system is to have lightweight code to speed up the response time. The former’s main task is to integrate the logic of the existing functional modules and make the integration platform through the flow of interfaces so that the workflow can flow in the functions of each module [25]. The latter is the management of the integration system alone, including the essential management work such as developing the process of personnel rights. In the HRMS database design, HRMS integrated database layer first considers the front Redis and other data caching applications to assist in processing data directly to the database layer to add, delete, and query operations and then design the system data corresponding to the type of database and storage devices. The selection of the final database implementation solution should consider the accuracy of the data required by the HRMS and ensure the consistency and integrity of the data.

6. Enterprise HRMS Test Implementation

6.1. Enterprise HRMS Testing

After the HR management system is designed, we need to test it, a process that has a vital role and will affect whether the system works properly. The primary purpose of this system is to unify and classify the data related to human resource management in the enterprise to reduce the difficulty of human resource management and improve efficiency. The HRMS must be highly secure and have a relatively stable working process. Testing of the system is often done to find deficiencies and errors in the design process so that they can be fixed. Therefore, we need to develop an accurate test plan to find potential problems in the system. The mark of a test’s success is its ability to find problems. Therefore, the main goal of this system testing is to find errors in the system and fix and improve them. There are two methods often used in system testing: white-box testing and black-box testing. This section uses black-box testing, which means testing the system without considering its internal environment and then watching the program run for any abnormalities. The system performance test pairs are shown in Figure 5.

The method of page switching is linked to the system’s responsiveness and Web performance, so it is necessary to try whether the page is switching speed and response time meets the usage requirements. In performing the test, it is essential to input specific error contents to see whether the system’s response speed and error response mechanism are complete. In addition, it is necessary to perform a unit testing process for the programs in the system. The code is the lowest level of the system components, so the code part needs to test whether the logical thinking is in line with the system functional requirements and the process and results of the system operation flow [26]. If there are explicit errors in the code, the testers can easily detect them, but a detailed inspection of the code is needed if there are implicit errors. Therefore, the code part needs to be checked and tested in detail. In addition, correctness testing of the system is also required. The program’s ability to perform the correct functions is essential in whether the system meets the user’s needs. Therefore, it is necessary to check the correctness of the procedures or variables called in the design and check the correctness of the data in the database. After passing the correctness test, the system functions can comply with the requirements. Performance testing is to test the system’s performance parameters, analyze the system’s performance parameters under a different number of requests by simulating the login and usage of the system, and test whether the system can meet the requirements in real business scenarios. We can discover hidden problems in the design through performance testing and eliminate them.

6.2. Multivariate Statistical Analysis of Enterprise Human Resource Management System Implementation

As the number of iterations increases, the scoring prediction error after gradient descent gradually decreases. Since DP-GD satisfies the differential privacy model by introducing random noise to the gradient, DPMF satisfies the differential privacy model by adding noise to the objective function; the noise introduced by both methods will carry data loss in gradient descent. The perturbation noise designed in this paper does not disturb the gradient data when performing gradient aggregation, so the accuracy of the comparison should be like that of the matrix decomposition model algorithm that does not do privacy protection and the scheme that uses encryption algorithm privacy protection [27]. It can also be seen from the experimental plots that the three methods have similar prediction error comparison results in two data sets with different amounts of data. Still, the algorithm’s error in this paper is better than the results of both after adding noise. This proves that this paper’s method can improve data utility, reduce the recommendation error brought by privacy protection, and provide an accurate recommendation service. A comparison of the prediction errors of the data set is shown in Figure 6.

An enterprise HRMS should have good robustness, not only in handling a small amount of work but also in increasing work. To prove the robustness of the HGAMW algorithm, several sets of corresponding comparison experiments are done in this section. The task completion time of HRMS is shown in Figure 7. When the workload is the same, the HGAMW solution has the shortest task completion time, followed by HGA, and the HEFT solution has the longest completion time. As the workload increases, the task completion time gets longer and grows slower when the workflow is small, and the task completion time grows faster when the workload is more significant. The execution time of both HGA and HGAMW proliferates with the increase in workload, and the execution time of the HGAMW algorithm is slightly shorter than HGA with the same workload. The overall efficiency of the HGAMW algorithm is higher due to the guidance of heuristic individuals.

BUGs in the system functions are classified into four categories according to the severity of the bugs. The most severe level is Class A bugs, which will lead to serious system crashes, causing the system to crash and affecting the whole system’s regular operation. Class D bugs are experiential problems of the system. After the corresponding functional tests, the system did not have any type A and B bugs. There were 11 and 14 bugs of types C and D, respectively. After detecting the bugs, the corresponding modules of the system have been repaired and revalidated through regression testing. So far, the method described in this thesis has realized the five modules of personal information management, work management, attendance management, reimbursement management, and entry/exit management, and completed a total of 17 detailed functional points in them. All the bugs found through the test have been solved. A comparison of before and after HRMS improvement is shown in Figure 8.

The design and testing of the human resource management platform can integrate the resources of professional talents belonging to an enterprise in an exclusive Web application system and, through the search function of the application system’s search engine, meet the supply and demand of professional talents in an enterprise, find out the personnel with relevant expertise in a short time, and learn the effect of the enterprise courses from the employees’ comments on the enterprise courses, to provide an essential reference basis for future corporate courses training content revision. In addition, there is objective data for reference on the performance of the participants in the training so that there is no disconnection with the reality and the interaction between the management and the employees can reach a balanced effect. This HR management system aims to become a professional talent cultivation management platform and then contribute to training a skilled regional workforce to create a professional and exclusive HR management platform.

7. Conclusion

In this paper, through the distributed environment of random matrix algorithm training requires the assistance of multiple terminals; to reduce the amount of computation, sensitive classification of items is performed according to the actual situation, and only sensitive items need to be scrambled during the training process, which not only improves the efficiency of the algorithm but also better preserves the utility of the data. In addition, dimensionality reduction technology is also used to compress the transmission gradient, significantly reducing the communication transmission overhead. Based on the architecture of the system, the algorithm is designed in detail in modules, the theoretical safety analysis and complexity analysis of the algorithm model are conducted, and the accuracy and time efficiency of the recommendation algorithm and the communication cost are verified on two real data sets with different dimensions. The critical technologies for developing the enterprise HR information management system are elaborated. This design uses MVC framework technology, B/S three-tier architecture, ORACLE database, and data mining for data anomaly detection to ensure the stable transmission of system data and improve the user experience. A black-box testing method was used for the system to simulate the working environment of the system and observe whether the code usually runs, and the test results were also shown. The test can prove that the system can achieve the expected goal and complete the function’s execution.

Enterprise human resource management is a step-by-step process that is interdependent with the enterprise’s strategic development and they affect each other. The current research theories and methods for HRM effectiveness evaluation are limited, and the scope of the HRM research design industry is limited. Due to our theoretical and practical deficiencies, the research in this paper has many imperfections and needs to be further explored in future research centers.

Data Availability

The data used to support the findings of this study are included within the article.

Conflicts of Interest

The author declares that there are no conflicts of interest or personal relationships that could have appeared to influence the work reported in this paper.