The Development Relationship between Cross-Border e-Commerce and Internet of Things Technology Coupling in Digital Economy Based on Neural Network Model.

Compared with the traditional trading model, the problem of information asymmetry in export cross-border e-commerce (CBEC) transactions actually exists. Since the buyers and sellers of export CBEC are located in different countries, it is difficult for both parties to make accurate judgments on each other's credit. Besides, the number of export CBEC financing is relatively small, and the financing situation is not optimistic. Fortunately, the Internet of Things (IOT) technology has a wide range of applications and strong social penetration, especially in e-commerce, which can effectively improve payment, logistics, and distribution. By employing the concept of the IOT, this paper expounds the application of the IOT in the e-commerce transaction process and the impact of the IOT on the development of e-commerce. Besides, this paper also proposes some security issues in the application of the IOT in e-commerce. In addition, this paper analyzes the relationship and development trend of CBEC and IOT technology; meanwhile, it also points out the practical problems currently faced by combining the regional advantage of a certain region. The e-commerce industry develops IOT technology and drives innovation-driven development in the region. In addition, this paper builds an evaluation index system for the development of export CBEC enterprises based on the analysis of the problems and reasons by using the sample data to train the BP neural network and using the test samples to test the model. It is found that the model accuracy rate is 89.47%. Then, this paper also takes the export CBEC company A as an example to conduct empirical research and to prove the operability of the credit evaluation model. Finally, it provides relevant suggestions for improving the credit evaluation system of export CBEC enterprises.

1. Introduction

Since 2015, the market scale of China's export CBEC industry has expanded rapidly, from 4.49 trillion yuan in 2015 to 8.03 trillion yuan in 2019, maintaining a relatively fast growth rate [1, 2]. In the scale of CBEC transactions, the proportion also far exceeds that of imported CBEC. With the rapid development of market size and profitability, the times also have new requirements for all aspects of China's e-commerce industry [3, 4].

On one hand, the credit evaluation of CBEC enterprises will help further promote commercial banks to improve their credit systems, the accuracy of their credit evaluation of CBEC enterprises, and the efficiency of loan issuance, which thereby can help to reduce their credit costs. On the other hand, it will help to further promote other evaluation institutions to improve their own evaluation systems and hence promote the sound and rapid development of the credit evaluation industry. Objective and effective credit evaluation results can also allow more investors to understand the company's situation more directly and quickly, which can bring more financing opportunities to the vast number of CBEC companies in urgent need of financing. Besides, it also can improve corporate financing efficiency and enrich corporate financing. Building an effective enterprise credit evaluation system can also supervise the enterprise itself. However, the current traditional e-commerce encounters development bottlenecks, the degree of automation is not high, the degree of scale is not perfect, and the remote support ability is not strong, which have become the main problems affecting the development of e-commerce.

It can be seen that in cross-border export transactions, enterprises have low awareness of credit. Industry supervision is not in place; that is, it is lack of relatively complete credit reward and punishment measures, which makes credit problems occur frequently and hinders the further development of export CBEC enterprises. In order to do a good job in the credit construction of export CBEC, the most important thing is to improve the export CBEC credit evaluation system, which can clearly show the credit level of the enterprise. Therefore, investors can easily obtain the credit results of the enterprise and understand the enterprise's credit rating. Credit strengths and credit weaknesses enable them to make scientific decisions. Besides, credit strengths and credit weaknesses are also of great help to the enterprise itself. The enterprise can adjust the enterprise structure according to the credit results, so that it can develop well and quickly. At present, most of China's export CBEC enterprises have credit problems, such as inferior products and integrity issues. The number of researches on credit evaluation of export CBEC enterprises is relatively small at present, and there is a business credit issues that are not taken seriously. Therefore, exporting CBEC is an important part of Chinese goods going global, and it is very important to further improve the credit evaluation system of Chinese exporting CBEC enterprises [5, 6].

As traditional international trade continues to weaken, CBEC driven by information technology and the Internet environment will surely become another application mode of the digital economy in the field of international trade. According to the “China SME Cross-border White Paper,” the current CBEC platform mainly assumes the role of matching the two sides of the trade. Meanwhile, it provides the online transaction process between the two parties, so the transaction data is available and large. With the continuous accumulation of transaction data, the interactive development of CBEC and the digital economy will surely drive the arrival of digital trade. Digital trade affects changes in transaction methods and production methods. Apart from being driven by the IOT technology, it also requires the coupled development of the economy and social environment [7, 8]. The rapid development of Internet of Things technology has brought about qualitative changes in people's lives. It is not only an opportunity for the innovation and development of the information industry but also has a significant and far-reaching impact on many fields and industries such as e-commerce.

At present, there is no uniform definition of digital trade in academia or even in the world. The US International Trade Commission defines digital trade as a transaction based on Internet technology that plays a key role in the process of ordering, producing, or delivering products and services. The organization divides digital trade into three categories: trade in goods, trade in digital services, and trade in data. In recent years, digital goods trade and CBEC led by China have been at the forefront of the world, while the United States has focused on the development of digital service trade and the free flow of data across borders [9, 10]. This paper believes that CBEC is regarded as a part of digital trade. In the initial stage, the role of e-commerce platform is to match the information of both parties and match transactions, and other transaction links are completed offline. Changes have taken place, and cross-border trade has gradually realized the online transaction process, and the online completion mode of the transaction process is also the main mode in my country at present; finally, with the gradual accumulation and precipitation of transaction data and the development of integration of digital technology, in addition, consumers' demand preferences are diversified, and it is urgent to mine massive transaction data to achieve precise matching of supply and demand [11, 12]. Therefore, CBEC will achieve qualitative changes under the continuous accumulation of variables in cross-border trade and evolve into digital trade. In general, CBEC is the initial manifestation of digital trade, and digital trade is an inevitable trend in the development of CBEC. Using the IOT technology, combined with the operational characteristics of international trade, can reduce the supply chain of trade goods and then quickly realize the flow of goods in international trade. A complete e-commerce information platform can bring users a good experience and promote the development of e-commerce informatization. Aiming at the problems existing in the construction of information platform in current e-commerce, the use of IOT technology is the most effective strategy at present. On the one hand, the combination of IOT technology and architecture can optimize the e-commerce information platform, which can effectively improve the intelligence level of the e-commerce information system, and can well promote the e-commerce information platform to meet the needs of information development. On the other hand, the use of IOT technology can strengthen the interaction between e-commerce platforms and consumers, timely feedback consumers' opinions and needs, and improve user experience, so as to attract more consumers and promote the effective development of e-commerce [13, 14]. With the rapid development of society and economy, people's life rhythm is accelerating, and the time to go out shopping is less and less. Coupled with the popularization of computers in people's lives, online shopping has become the most convenient and popular shopping method at present, and the e-commerce industry has also become the commercial industry with the most development potential in the 21st century. After successful online shopping, consumers obtain goods in the form of express delivery through logistics companies, which requires the construction of logistics systems in the development of e-commerce, and the introduction and development of IOT technology can help e-commerce build an intelligent logistics system [15, 16]. Once the logistics company signs and receives the package, it can affix the RFID label with the package information to each package and transmit the package information to the Internet, so that consumers can track the information of their own packages anytime and anywhere, so that the information can be collected. Transparency guarantees consumers' security needs, improves consumers' trust in e-commerce, and is conducive to the sustainable development of e-commerce [17, 18].

China's e-commerce information platform started relatively late, has no rich experience, and has many problems. First of all, the sharing and transmission of e-commerce information is separated from reality, mainly because the actual operation process of e-commerce is not fully considered in the design process of the information platform architecture [19, 20]. The second is the lack of value-added services, resulting in the current Chinese e-commerce information platform unable to adapt to the development needs of the IOT technology. Finally, the operation effect of the e-commerce platform is low, and there is no efficient channel, mainly because there is no online and offline connection with the information platform in actual operation. At present, there are many technical problems in e-commerce under the IOT platform, such as chips, equipment, platform software, and hardware, which are far from other countries. Cloud computing currently has no perfect data processing model and the relationship between IPv4 and IPv6. The exchange of information between them is not yet possible. At the same time, the IOT e-commerce has not formed a mature operation plan, and it lacks competitiveness in cost control, which makes its commodity prices high, and consumers do not have enough ability to consume, resulting in a weak market [21, 22]. In addition, irregular policies and regulations, lack of security and privacy protection, and inconsistent technical standards are all problems that arise in the development of e-commerce under the current IOT technology [23, 24]. In the CBEC logistics system, there is also a complex nonlinear relationship between various risk indicators, and the genetic neural network is a mapping of a highly nonlinear relationship [25]. The neural network model optimized by the genetic algorithm can objectively evaluate the risk. BP neural network is an intelligent algorithm established by mathematical methods to process and memorize information by simulating the neural network structure of the brain. It constructs various neural networks with different topological structures according to neurons, so as to realize the simulation and approximation of the researched things. It can self-learn, self-organize, and fit arbitrary nonlinear functions, so it has a wide range of applications.

Therefore, based on the BP neural network, this paper studies the development relationship between the CBEC of the digital economy and the coupling of the IOT technology. Starting from the concept of the IOT, the article expounds the application of the IOT in the e-commerce transaction process and the impact of the IOT on the development of e-commerce and puts forward some security problems in the application of the IOT in e-commerce, so as to give some suggestions according to the problems. The application of this technology by e-commerce enterprises under the Internet of Things technology is not only an innovation of technology but also an innovation of enterprise management mode. It can break through the limitations of traditional e-commerce and conduct information interaction more efficiently and directly.

2. BP Neural Network

Through CBEC logistics risk research, it is beneficial to promote effective risk aversion between enterprises and achieve cooperation benefits; it is beneficial to control unnecessary risk losses, missed opportunities, etc., and promote all aspects of physical logistics and CBEC lines. In addition, the operation of CBEC logistics is actually a process of collaboration and competition among various stakeholders. Through research on risks and sorting out the internal mechanism of each link, the risk management of CBEC logistics among enterprises will be more clear, and it will provide a corresponding reference for the government to further improve the CBEC logistics supervision system. Business enterprises and e-commerce platforms give relevant management inspiration. At the same time, it will further promote the control and risk aversion of information flow, capital flow, and logistics. The application of BP neural network, according to the characteristics of CBEC logistics risks and the research results of scholars, selects genetic algorithm to optimize the initial weight threshold of neural network and establishes CBEC logistics risk evaluation model and algorithm, which can effectively predict and analyze the CBEC logistics risk. The logistics risk of overseas e-commerce is minimized.

The artificial neural network operates according to the algorithm and continuously trains and learns by processing the original data to form a certain rule. After the training is completed, the original input value is reentered, and the value closest to the expected output value will be obtained. The BP algorithm includes signal forward propagation and error back propagation; that is, the direction of input to output is used when calculating the error output, and the direction of output to input is used when adjusting the weights and thresholds. During forward propagation, if there is an error between the actual output value and the expected output value, the error backpropagation will be performed, and the weights and thresholds will be continuously adjusted through the error backpropagation, and the corresponding parameter information with the smallest error will be finally determined. The BP neural network has a strong learning ability in the study of nonlinear events. Its main principle is the gradient descent method. The search target sample satisfies that the error mean square error between the actual output result of the training network and the expected output is the smallest; that is, the optimal result is obtained.

BP neural network is a multilayer feed-forward network, which is characterized by training according to the back propagation of errors, and by using the gradient search technology, the result of minimizing the error of the calculation results is achieved. The specific structure of BP neural network is shown in Figure 1.

Figure 1

The specific structure of BP neural network.

Let the BP neural network structure be where n, q, and m are the dimensions of the network.

The network includes the weight from the ith neuron in the input layer to the jth unit in the hidden layer, which can be expressed as where w is the weight.

The weight from the jth neuron in the hidden layer to the kth neuron in the output layer can be expressed as

The steps of the BP neural network learning algorithm are as follows (Figure 2):

Figure 2

The steps of the BP neural network learning algorithm.

Initialization: set the initial values of the weights and thresholds of the network to values in the interval [0, 1]

Forward propagation of the network: let the input of the pth group of data samples be

where x is the variable.

Expected output is

where t is the output.

Then, the output information of the jth neuron in the hidden layer is

The hidden layer passes the output information to the output layer, and the final output is as follows:

where y is the final output

(3) Calculate the squared error E of the BP neural network: let the actual output of the pth group of samples be

Then, the network error squared sum E can be expressed as follows:

Judge whether the sum of squared errors E converges to the given learning accuracy ε, if E ≤ ε, the algorithm ends, and the network stops training; otherwise, go to step (5)

(4) Error back propagation: starting from the output layer, backpropagation layer by layer, using the steepest descent algorithm in nonlinear programming, modifying the weights of the network according to the negative gradient direction of the error function E, and adjusting the weights of the network layer by layer

In the formula, η represents the step size or the network learning rate. The introduction of η is to speed up the convergence rate of the network. In order to increase the stability of the network learning, a momentum parameter a is usually added to the weight correction formula, then the nth learning. The modified formula of the weight is

(5) Repeat steps (3) and (4) until the output error of the sample meets the predetermined condition, and the network training ends: the BP neural network algorithm is combined with the genetic algorithm, and the genetic algorithm is used as an auxiliary to optimize the connection weights between the layers of the neural network. Establish a mathematical model of cross-border e-commerce logistics risk assessment based on genetic neural network

3. Construction of CBEC Logistics Risk Evaluation Index System

The construction of the CBEC logistics risk evaluation index system is based on the identification of potential risk factors in the entire chain of CBEC enterprises in the operation of the logistics system and then overall consideration and analysis to obtain scientific and reasonable risk evaluation indicators. The construction of the risk assessment index system should follow the following principles.

3.1. The Principle of Scientific Objectivity

The establishment of a CBEC logistics risk evaluation index system is based on a scientific theoretical basis and requires a large number of theories and research methods to support it, so that it can be objective and true. Cross-border logistics risk assessment needs to be carried out in different environments and under different subjects, with the participation of people from different industries, and is vulnerable to external interference. Adhere to a scientific, objective, and reasonable position to conduct scientific and objective risk assessment.

3.2. Hierarchical Progressive Principle

First of all, if risk factors exist or occur according to the different business activities of each link, it is necessary to split the CBEC logistics first to obtain a set of risk factors and then select appropriate indicators according to the principle of integrity, so that they can effectively reflect the CBEC logistics. The uniqueness of CBEC logistics can also make the entire risk index system not interfere with each other. The full consideration of multiple logistics nodes and multilevel analysis and selection makes it effective and characteristic and can cover the entire CBEC logistics system.

3.3. Practical Principles

The establishment of the CBEC logistics risk evaluation index system is not aimed at a certain enterprise but has a commonality that enables it to be effectively applied by other CBEC enterprises in the industry. The availability of index data should also be considered when establishing the index system. The availability, effectiveness, and sensitivity to risk of index data should be considered when establishing index system. Thus, we can make it more practical.

3.4. Comprehensive Comprehensiveness Principle

The risks of CBEC logistics should be taken into account both internally and externally. Different countries and external factors involved in CBEC integrated logistics should be considered comprehensively by seeking common ground while reserving differences. Therefore, the indicator system should integrate the commonalities of various risk factors but at the same time cannot be generalized.

The assumptions of this model are as follows: select indicators from the characteristics of custom clearance risk, platform risk, process risk, organizational risk, and environmental risk in the overall operation of CBEC logistics, as well as the generality of different cross-border logistics models to CBEC logistics. The commercial logistics risk evaluation index system is initially constructed. The entire risk index system consists of five aspects: platform risk, organizational risk, custom clearance risk, process risk, and environmental risk. This paper chooses the method of combining qualitative and quantitative indicators. Quantitative indicators are more scientific and rigorous, and fewer qualitative indicators reduce the influence of subjective factors to a certain extent. Each risk index directly or indirectly evaluates the risk. Starting from the entire logistics chain, there are risk factors in each link. It is necessary to select the influencing factors that can describe the risk index in data and to comprehensively measure and comprehensively qualitative and quantitatively characterize descriptions simultaneously. There are directly influencing factors as well as the comprehensive and indirect factors affecting risk indicators. Each risk index is representative, so it can comprehensively evaluate the logistics risk of CBEC. In the custom clearance risk, it is assumed that the product custom clearance efficiency in the measurement index is calculated according to the actual custom clearance time of the product. Due to different custom clearance countries and different products, the product custom clearance efficiency cannot be comprehensively measured, so it is recommended to exclude it. The process risk mainly includes two parts: logistics and warehousing. According to the actual logistics, transportation, and distribution situation, it is recommended to add the batch rate and transfer rate at the end of distribution. This indicator can effectively reflect the efficiency of logistics and transportation and the effective times of loading, unloading, and sorting and also has a certain impact on risk evaluation. Organizational risk mainly recommends adding service satisfaction rate and effective complaint rate for service ability risk. This indicator mainly refers to the satisfaction level of logistics services provided to customers, which can intuitively reflect the quality of logistics services, and the influence of logistics service brands in the market and business development capabilities is the reflection of the risk impact of the corporate brand in the market environment.

4. Empirical Research

Before training the BP neural network, the relevant input and output values must be mastered in order to perform normal training and learning. The output value refers to the initial credit rating data of the sample. This paper selects the basic data and the expected value of credit level of representative Chinese export CBEC enterprises (self-operated export CBEC enterprises and platform-based enterprises) to train the model. Considering the availability and authenticity of the data, this paper selects the export CBEC enterprises listed on the main board, the new third board, the small- and medium-sized board, and the ChiNext as the data sample. BP neural network is an intelligent algorithm established by mathematical methods to process and memorize information by simulating the neural network structure of the brain. It constructs various neural networks with different topological structures according to neurons, so as to realize the simulation and approximation of the researched things.

Before constructing the BP neural network model, the BP neural network should be trained and learned with sample data, and the samples used for training and learning are training samples. In this paper, the input node is 32, and the output node is 1. The model is trained and debugged many times according to the previous hidden layer node empirical calculation formula. The model is trained best when the number of hidden layer nodes of the model is 10.So the final BP neural network structure is 32-10-1. Input the set training code into the MATLAB software; train 148 groups of training samples; continuously adjust the weights, thresholds, and the number of hidden layer nodes according to the training results; and finally get the best running effect. The comparison between the actual output value of the training sample and the input value of the original credit rating is shown in Figure 3, and the fitting effect is shown in Figure 4. As shown in Figures 3 and 4, the accuracy of the training model can be clearly seen, and the accuracy of the training samples has reached 100%, indicating that the training effect of the data in the BP neural network model is significant.

Figure 3

Comparison between the original output value of the training set and the real value.

Figure 4

Regression fit.

Firstly, genetic operations are carried out. After continuous crossover, mutation, inheritance, etc., in the sample population, the target sample individual with the greatest fitness is searched; that is, the initial weights and thresholds of each layer connection of the optimized neural network are obtained. The optimization results are assigned to the neural network, and the optimized neural network structure model is obtained after training sample data. Bring the data values of each risk measurement index of CBEC logistics into the model for training, use the data samples to first learn the nonlinear function relationship, and then use the control samples to verify and obtain the final CBEC logistics risk evaluation output value, which is the risk of CBEC logistics in this example. According to the level interval of each risk rating in the model, the risk level of CBEC logistics risk is determined by comparing with the final CBEC logistics risk value. The training results are shown in Figure 5, and the result fitting (Figure 6) is as follows: combining Figures 5 and 6, the actual output of the training sample and the expected output value have little difference, the fitting is well completed, and the parameter values of the neural network are further adjusted. It is again verified that the model has a good ability to learn nonlinear functional relationships.

Figure 5

Training results.

Figure 6

Result fit.

The sensitivity analysis is carried out on the CBEC logistics risk evaluation index. On the basis of the trained genetic neural network, the sensitivity analysis is carried out on the parameters of each evaluation index in the CBEC logistics risk of enterprise A.

Knowing its risk evaluation result, reduce the value of each risk index in the network by 0.05, and keep other setting parameters unchanged. Run the operating software to obtain the risk evaluation result. When the value of any evaluation index changes, the risk evaluation result corresponding to the model is shown in Figure 7. Besides, it is also compared in Figure 8.

Figure 7

Genetic neural network risk assessment results after each index value changes.

Figure 8

Comparison.

According to the changes of the output results, it can be reflected that the changes of the index values have little impact on the output results, and the overall sensitivity when changing the index values is low. The results of enterprise A's CBEC logistics risk evaluation, combined with the fluctuations and changes of the measurement indicators, 5-dimensional risk factors, custom clearance risk, platform risk, process risk, organizational risk, and environmental risk, correspond to 21 measurement indicators. Among them, the sensitivity of process risk measurement indicators fluctuates more than other risk measurement indicators, so enterprise A is in process risk. The transportation risk is mainly due to the long transportation distance, the number of transshipments during transportation, the combined transportation of various transportation vehicles, etc., which are the direct reasons for the high process risk. From the perspective of indicators, improve and strengthen transportation and distribution efficiency, transportation equipment utilization rate, etc. In terms of warehousing risks, it can effectively reduce the risk value; in the warehousing risk, combined with the warehousing situation of company A, its warehousing business provides overseas warehouses, bonded warehouses, etc., and for indirect transit warehouses, choose to rent third-party warehouses or build simple transit warehouses, transit stations, etc. Although there is a certain reduction in warehousing costs, its management and warehousing resource utilization directly affect the warehousing risk value. From the analysis of the results of the example, it can be concluded that warehousing resource utilization and warehousing facility management should still be the main breakthroughs to control warehousing risks. By analyzing the risk evaluation results according to the example model in organizational risk and platform risk, it can be concluded that company A has achieved relatively mature management, control, and integration between companies in terms of organizational risk. Platform risks are mainly aimed at computer technology, etc. The current results reflect well, and for the better development of enterprises, continuous breakthroughs and research in this area are still needed. The risk is analyzed according to the results of the case of enterprise A, so as to effectively control and improve the logistics risk of CBEC as a whole.

Based on the evaluation results of enterprise A's CBEC logistics risk and the research on CBEC logistics risk in this paper, CBEC logistics enterprises involve transportation and warehousing (overseas warehouses or bonded warehouses, etc.) and have independent CBEC logistics. The evaluation method of this paper can be applied according to the business modules involved in the business platform and the cooperation of many enterprises. The CBEC logistics risk evaluation model based on genetic neural network can give relevant evaluations for the risks of many CBEC logistics enterprises. According to the evaluation results, the effective management and control of its risk matters and further risk research are used as a reference.

5. Conclusion

BP neural network has certain advantages in credit evaluation of export CBEC. The BP neural network method is used to evaluate the credit of export CBEC enterprises, and the model calculation accuracy is high. The BP neural network model itself is relatively stable, the operation speed is relatively fast, and it has advantages in processing a large amount of data, reducing human subjectivity, and being more standardized and objective. This paper uses the BP neural network to train and learn the sample data. The accuracy of the credit evaluation results is good and at a high level, which can provide scientific reference for investors and managers.

The BP neural network selected in this paper is a method suitable for a large amount of data. As the number of samples increases, its accuracy will also increase. The number of samples can be continuously increased in future research to improve the accuracy of the model.