Analysis of the integration and application of artificial intelligence Internet of Things technology in the construction mode of innovation and entrepreneurship education

Advancing artificial intelligence and the Internet of Things in this new age has infused innovation and entrepreneurship education with fresh content, innovative ideas, theories, technologies, and distinct personalities. Objectives and principles in the realm of innovation and entrepreneurial learning. The Internet of Things is crucial in the emerging wave of information technology. As indicated by its title, “The Internet of Things serves as the web linking various entities”.Intelligent sensors are employed to create the Internet of Things node for data collection in managing innovation and entrepreneurship resources within academic institutions. Through studying the current situation of innovation and entrepreneurship education resources in colleges and universities, we understand the needs of management, build a sound innovation and entrepreneurship education resource management model based on Internet of Things technology, improve the level of innovation and entrepreneurship education resource management, develop artificial intelligence, and build a new pattern of innovation and entrepreneurship education, Creative entrepreneurship education is effective, unified and complementary, which can jointly cultivate innovation and the development of real talents[1]. See Figure 1 for the Internet of Things framework and AI innovation and entrepreneurship education.

Education in innovation and entrepreneurship through the Internet of Things expands upon managing such resources, primarily focusing on utilizing the smart Internet of Things. Within different sectors, particularly in innovation and entrepreneurial learning among diverse businesses and establishments, encompassing colleges and universities. As the primary entity for education in innovation and entrepreneurship, it can instantly comprehend the fundamental aspects of such education and facilitate enhancements and modifications via digital and interactive connections[2-4].

Conversely, it can also promptly identify possible gaps in the innovation and entrepreneurship education process. The steadiness of resources related to innovation and entrepreneurship education influences the robust growth of the environment for innovation and entrepreneurship. Consequently, creating a practical and efficient management of resources for innovation and entrepreneurship education is especially crucial. Refer to Figure 2 for the assessment and analytical structure for the Internet of Things, AI innovation, and entrepreneurship. Educational materials for ships.

The widespread integration of the Internet of Things with artificial intelligence enhances sensor placement and external interference resistance in the Internet of Things innovation and entrepreneurship education hub. T Data from the Internet of Things regarding innovation and entrepreneurship education resources serve as exemplary data for examining the current state of these educational fields. Utilize the BP neural network and decision tree algorithm during the learning phase to isolate sample data characteristics and scrutinize state categorization, aiming to develop the innovation and entrepreneurship education resource management m model[5].

2

Build an evaluation model of entrepreneurial education resources based on artificial intelligence technology

2.1

Internet of Things management model of innovation and entrepreneurship education resources

The innovation and entrepreneurship education resource management system designed in this paper, based on the CoAP protocol of the Internet of Things, mainly provides two interfaces. One is to map to the CoAP RESRful interface[6], enabling the CoAP client in HTTP Web Service to request response data. One is that it is unnecessary to map to the CoAP RESTful interface. Usually, the data related to the database is read and processed in HTTP Web Series, and then the data is requested accordingly. The interfaces of these two methods provide HTTPRESTful API for innovation and entrepreneurship education resource management servers.

The Internet of Things service platform based on the CoAP protocol will provide interfaces that do not need to be mapped to CoAP RESTful APIs, such as data storage, data analysis, user management, sensor data extraction, and other business interfaces. The sensor data extraction will include historical sensor acquisition data, sensor information data, etc. When the innovation and entrepreneurship education resource management server calls the HTTP RESTful API provided by the IoT service platform, the IoT server platform will first use routing mapping to call the corresponding service, and the service business logic will read the corresponding data from the database, then make an ORM mapping to the object SensorInfo, and finally return the object Sensor data format as JSON to the service caller[7].

Among them, we use Spring Data MongoDB database access driver to achieve interaction with MongoDB. MongoDB is an important collection concept. The part that groups and stores data is called a collection. MongoDB provides a unique ID for such a collection. We only need to use the application. The Spring Data MongoDB database driver can be integrated with the Spring framework by configuring in XML[8].

We configure all Action classes in struts. XML in the struts two frameworks The class attribute in the XML file is specified by the container bean ID in the Spring MVC framework so that the container of the Spring MVC framework manages the Action object. The container in Spring MVC starts from the application. Get the specific implementation of the bean in the XML file. The Action class contains the gas business implementation logic. There are roughly three interfaces to the persistence layer. One is the data storage interface generated by the integration of the Mybatis framework and Postgresql, the other is to call the HTTP RESTful API of the Internet of Things service platform, and the last is the third-party interface, such as the payment interface and the SMS interface.

Resource managers in innovation and entrepreneurship education can be categorized into various user roles, each utilizing distinct functional points within the system. According to the actual demand survey, it is not necessary to implement content permission management. The so-called content permission refers to the same type of data. According to different user permissions, the range of data that can be queried and edited differs. If there is a demand for data content permission in the future, it can be achieved by subdividing the function permission under less complex circumstances. All colors are horizontal, and there is no inheritance structure; A user can only belong to one role, not multiple roles. You can add new roles if there are requirements for role inheritance and multiple roles. During the system design and implementation, the addition of roles should be strictly controlled to avoid role flooding. The permissions section defines the systems, functions, roles, functions, and users. This information constitutes the basic information of the permission part. Other systems and functions must obtain user and permission information through the permission interface[9].

In this research, the innovation and entrepreneurship resource node’s sensing layer employs nodes with multiple sensors for information focus. Utilizing the Internet of Things’ system architecture, the project employs a two-dimensional code label linked to the Fixed asset devices gather gathered data into the fixed asset center’s database via wireless network communication technology and achieves the gathering, fixing, and borrowing control of innovation and entrepreneurial education resources via the backend business management system.

Web services are the basic building blocks for distributed computing on the Internet and an important part of SOA architecture. This project adopts the SOA architecture, which brings the following main benefits: (1) It realizes a physical three-tier structure. (2) The external interfaces of all domain models are WS, so they can be accessed by various terminal applications, which improves the repeatability of interface use and the return on investment. (3) Plug and play of the application will further expand and prepare for the school’s application. (4) it can be encapsulated in the domain model without large-scale software modification in the face of changes. Web Services provide services to smart clients or other systems in the security system framework through HTTP or other network protocols[10].

Apache CXF (Celtix+XFire) is an open-source framework for developing Web services. A range of transport protocols, including HTTP, JMS, and JBI, can operate these Web services, which are compatible with diverse protocols like SOAP, XML/HTTP, and RESTful HTTP. Refer to Figure 3 for a sample model illustrating innovation and entrepreneurship. Managing educational resources related to the Internet of Things.

2.2

Filtering algorithm for innovation and entrepreneurship education resources of the Internet of Things

By gathering daily data on business operations related to innovation and entrepreneurship education in real-time via the widespread Internet of Things, one can assess the effectiveness of these educational resources. We are actively enhancing the effectiveness of teaching innovation and entrepreneurship. The significance of the Internet of Things in overseeing and tracking resources for innovation and entrepreneurship education is evident in its crucial role in both the development and implementation of these educational areas. In the context of IoT data gathering and handling for educational resources in innovation and entrepreneurship, the node’s signal is prone to disturbances from external environments. Enhancing the effectiveness of mapping is of paramount importance. Determining the precise location of the node through the application of appropriate algorithms. The study employs the enhanced LANDMARK positioning algorithm to enable IoT nodes to withstand external environmental interference and ensure precise positioning.

The supplementary reference point label remains constant, with its signal intensity readily altered by external factors, impacting the precision of its positioning. The aim is to enhance the dependability of the correlation between the RSSI value and the actual p. During the offline database construction phase; the initial signal is repeatedly gathered at every reference point, followed by filtering preprocessing to derive a consistent and dependable auxiliary positioning RSSI signal value. Addressing this issue, the system utilizes the Gaussian average filtering technique to remove outliers exhibiting significant fluctuations and deviations. This filtration technique offers reduced computational effort, rapid processing, and significant efficiency effect. The procedure proceeds thus: Following the setup of the LANDMARC positioning system, every AP obtains the signal strength value vector Em from all reference tags. Subsequently, m sets of RSSI values and the top ten groups are acquired for every reference tag. Merge into a collective for calculation. The highest and lowest values are eliminated to determine the stable mean value. Subsequently, the mean value of the signal sample is assessed along μ with its standard deviation σ. As shown in Formulas (1) and (2). (1) $μ = \frac{1}{m} \sum_{i = 1}^{m} {RSSI}_{i}$ (2) $σ = \sqrt{\frac{1}{m} \sum_{i = 1}^{m} {(R S S I_{i} - μ)}^{2}}$

Finally, calculate the mean μ ☆ and original variance σ. The fingerprint information F as the reference label is stored in the fingerprint database. F is shown in Formula (3), where i∈(1,m), l∈(1,n). (3) $F_{i} = (x_{i}, y_{i}, μ_{1}, μ_{2}, \dots, μ_{l}, σ_{1}, σ_{2}, \dots, σ_{l})$

Within the traditional LANDMARC positioning algorithm, the KNN algorithm is typically employed during the online positioning phase. The initial k smallest values in vector E are picked to approximate the coordinate data upon choosing the suitable K value. Choosing the K value significantly impacts the precision of the LANDMARC positioning system. Typically, the precision in positioning attains a significant degree when k equals 3, 4, or 5. Upon the tag’s emergence in the indoor setting with an identical likelihood, the ideal k value selection is directly influenced by how the reference tag is deployed. In this chapter, the reference labels are set up within a triangular network; hence, the choice of k=3 enhances the system. Once the least three values in the E vector are chosen, position the coordinate points of the three reference labels at the center of t. the circle, and form three circles, each with a radius equal to 4/5 of the distance between the centers of two circles. Opting for 4/5 aims to enhance the likelihood of intersecting three and two circles. The algorithm steps are as follows: 1)

Ascertain the trio of reference label coordinates (x₁,y₁) (x₂,y₂) (x₃y₃), along with an extra fourth coordinate (x₄,y₄)

2)

Determine the gap between (x1,y1) and the remaining two reference points, noting it as d₁₂, d₁₃, and d₁₄ in that order. Choose four-fifths of the peak value as the radius and document it as R₁; similarly, R₂R₃ is computed symmetrically. (4) $R_{1} = Max (d_{1 i}) = Max \sqrt{{(x_{1} - x_{i})}^{2} + {(y_{1} - y_{i})}^{2}} i = 1, 2, 3$

3)

The coordinate equation of three circles: (5) ${(x - x_{i})}^{2} + {(y - y_{i})}^{2} = R_{i} i = 1, 2, 3$

When the following formula: (6) $\begin{array}{rcl} \sqrt{{(x_{1} - x_{2})}^{2} + {(y_{1} - y_{2})}^{2}} \leq R_{1} + R_{2} \\ \begin{array}{l} \sqrt{{(x_{2} - x_{3})}^{2} + {(y_{2} - y_{3})}^{2}} \leq R_{2} + R_{3} \\ \sqrt{{(x_{1} - x_{3})}^{2} + {(y_{1} - y_{3})}^{2}} \leq R_{1} + R_{3} \end{array} \end{array}$

If the above formula is not satisfied at the same time, the following can be obtained: (7) $S R e f x = \frac{x_{1} + x_{2} + x_{2}}{3} S R e f y = \frac{y_{1} + y_{2} + y_{3}}{3}$

At the same time, it is proved that three circles intersect each other: 4)

On the premise that three circles intersect each other in 3), the simultaneous equation: (8) $\begin{matrix} x = R_{1} \cos θ + x_{1} \\ y = R_{1} \sin θ + y_{1} \\ {(x - x_{2})}^{2} + {(y - y_{2})}^{2} = R_{2}^{2} \end{matrix}$

5)

Find the midpoint of two intersection points and record it as M12:

6)

Similarly, we can find the midpoint M13 of the two intersection points of circle x1 and circle x3, and the midpoint M23 of the two intersection points of circle x2 and circle x3: (9) $\begin{array}{l} M_{13} : (\frac{x_{13}^{1} + x_{13}^{2}}{2}, \frac{y_{13}^{1} + y_{13}^{2}}{2}) \\ M_{23} : (\frac{x_{23}^{1} + x_{23}^{2}}{2}, \frac{y_{23}^{1} + y_{23}^{2}}{2}) \end{array}$

Then, write according to the three midpoints of M₁₂M₁₃M₂₃: (10) $\begin{matrix} \hat{x} = (x_{M_{12}} + x_{M_{13}} + x_{M_{23}}) w \\ \hat{y} = (y_{M_{12}} + y_{M_{13}} + y_{M_{23}}) w \end{matrix}$

3

Internet of Things Innovation and Entrepreneurship Education Resources Internet of Things Management Data Artificial Intelligence Analysis Model

The data of each sensor data node of the Internet of Things is obtained. After filtering the environmental interference clutter, it is necessary to analyze and optimize the massive data obtained from the data obtained from the innovation and entrepreneurship education resource management node of the Internet of Things. For the data analysis of the Internet of Things node, it is necessary to start from three aspects: classification, regression, and aggregation. This research applies AI to the data analysis of IoT nodes mainly by abstracting the actual problems into corresponding mathematical models and decomposing them into the organic combination of three tasks: modeling and solving. See Figure 4 for the AI IoT node sensor data analysis process. 1)

Innovation and entrepreneurship education resource management Internet of Things node artificial intelligence analysis VGG model

The advanced learning model of VGG is called the visual geometry group, which can be regarded as the advanced version of AlexNet[11]. It faces recognition and other tasks. Initially, the algorithm model was mainly aimed at elaborating on the relationship between AI networks and the accuracy of large-scale classification. The results of scholars’ running research on the VGG model show that the AI system is positively correlated with the network generalization ability and negatively correlated with the error rate. The VGG model primarily consists of a fold, pool, and solid connection layer. The primary technique for determining the folding layer involves computing the folding layer[12]. Typically, the process of folding necessitates factors like the size of steps and the extent of padding. Once the zoom function is finalized, the activation feature is typically implemented. The activation feature aims to enhance network data’s comparability by incorporating nonlinearity. These active functions are typically employed (ReLU, Sigmaid, ThanH). The equations are outlined as(11), (12) and (13). (11) $h (z) = \max (0, z)$ (12) $h (z) = \frac{1}{1 + e^{- z}}$ (13) $h (z) = \tanh (z)$

In addition, after extensive convolution application checks to input characteristic figures for convolution operation, it is also necessary to filter the information of characteristic figures. Pooling mainly uses the maximum value and average value for the operation. The full connection layer is mainly used to fuse feature information. It usually connects the full connection layers in the order of dimension reduction[13]. The final application of this layer is a nonlinear combination of characteristics, lack of feature extraction, and learning ability. In graphic Analysis, small-size convolution kernels are used to reduce parameter analysis pressure. 2)

Google Net Artificial Intelligence Network Model

Google Net model has designed a special network structure (sparse network structure), which makes the network performance less affected by the network depth and reduces the number of parameters. GoogLeNet, intended as the 22nd layer, features parameters constituting a twelfth of AlexNet and a fourth of VGG[14]. Comprising a compact folding core and a 3x3 pool layer, the network structure is characterized by its sparsity. Not only does this framework enhance network efficiency, but it also safeguards the effectiveness of computational resources. AI sparse architecture is shown in Figure 5. 3)

Transformer Deep Learning Model

Transformer is gradually applied to image processing tasks to enhance the correlation between image feature map pixels and increase the global receptive field. The attention mechanism is a key component of the Transformer, which focuses on the part of input data with more feature information. The calculation steps of the self-attention mechanism are as follows[15]:

convert data samples into embedded vectors;

Randomly initialize W_q, W_k, and W_v weight matrices to calculate q query vector with input vector, k key vector, and v value vector;

Calculate the score of each q vector and k vector through convolution. The formula is as follows (14). (14) $score = q \cdot k$

To stabilize the gradient, divide the calculated score by $\sqrt{d_{k}}$ , where d_k is the dimension of the bond vector;

Use the SoftMax function to calculate the score and get the weight of the v vector;

Weighted sum of vector v and output vector z. A feedforward neural network links in sequence through two complete connection layers, typically employing ReLU or linear activation functions, primarily to reduce data dimensionality. The specific formula is as follows (15)[16]. (15) $F F N (z) = \max (0, z W_{1} + b_{1}) W_{2} + b_{2}$

Where z is the vector z output by the self-attention mechanism, the transformer structure of AI is shown in Figure 6.

The organization has limited resources and cannot improve education as much as possible. Maximize the use of educational resources to match each investment resource while considering the complexity of various resources that are directly related to educational results to effectively allocate each resource to the most appropriate direction.

Efficiently utilize resources to foster innovation and initiative. The assets of tertiary education and universities, along with the cultivation of innovative and entrepreneurial mindsets, are influenced by various elements, and their allocation is not linear. The ultimate aim is to Enhance educational efficiency by elevating the overall quality of education[16]. (16) $U = \frac{\sum_{j = 1}^{m} μ_{j} O_{j}}{\sum_{i = 1}^{n} φ_{i} I_{i}}$

The effectiveness of allocating resources in the realms of creativity and initial education. The university’s creative education resources input-output index system consequently establishes student-faculty ratios, the count of new administrative teachers, and the number of teachers with a business background outside China; educational tools and apparatus hold China in high regard, encompassing the area occupied by practice platforms, the China educational base, special funds, among seven indices of efficiency in allocating resources for innovation entrepreneurship education[17]. (17) $\begin{array}{l} A_{k i} = \frac{X_{k} + Δ X_{k}}{S_{k}} (k = 1, 2, \dots, K) \\ s . t . A_{k i} < 1 (i = 1, 4) \\ 0 ⩽ A_{k i} ⩽ 1 (i = 2, 3) \\ A_{k i} > 1 (i = 5, 6, 7) \end{array}$

Within this group, Sk stands as the most populous university student at KTH; X_k and ΔX_k symbolize the fair allocation of various educational aspects of creativity and startup among students at KTH University, respectively.

Considering the objective function comprehensively, the multi-objective optimization function of resource allocation for creativity and startup education is[17]. (18) $\max U_{k} = \frac{\sum_{j = 1}^{m} μ_{j} O_{j}}{\sum_{i = 1}^{n} φ_{i} I_{i}}$ (19) $\max F_{k} = \sum^{n} φ_{i} A_{k i}, k = 1, 2, \dots, k$

The PSO algorithm represents an innovative approach to swarm intelligence, grounded in the principles of ant colony algorithm and heuristic optimization. Its features include notable stability, rapid convergence, straightforward coding, and user-friendly programming. An AI algorithm focusing on multi-objective optimization, utilizing particle swarm optimization, has been developed to fulfill the demands for extensive enhancement of multi-objective optimization and search methodologies[18]. (20) $\begin{matrix} V = w V_{i} + c_{1} r {()}_{1} (y_{i} - P_{i}) + c_{2} r {()}_{2} (g_{i} - P_{i}) \\ P_{i + 1} = P_{i} + V_{i + 1} \end{matrix}$

A neural network is a nonlinear mathematical simulation in system control theory. It uses simple nonlinear operations to define complex expressions. This can not only objectively and effectively evaluate evaluation results influenced by human factors but can also change the complexity of traditional evaluation methods. In this model, the learning and evaluation system is designed as a reference parameter of the BP nervous system[19-22]. The logical framework of the evaluation system and teaching examples are developed. By default, each group of control data for training is entered into the network system, and the evaluation results are regarded as the starting factors of the neural BP system.

If the required output layer is not reached, the counter current is in the wrong direction of the rotation signal. By changing these two processes, the vector space realizes the gradient reduction strategy of the error function and the dynamic, iterative search of the weight vector group to minimize the network error function, thus completing the information retrieval and storage process. This means that the sample input data at the input level, calculated using the transfer function and compared with the expected data, can be output[22-25]. If an error occurs, return it to the top level to change the weight of the vertex.

Different network structures are endowed with different problem-solving capabilities in the AI algorithm. It can increase the number of nonlinear meshes learning ability of neural networks. The neural network supports BP layer, so a three-level BP neural network can supplement any N-level m-Size diagram. In the assessment model of the creativity and startup quality education based on a neural system, this paper applies a three-level BP neural network structure hidden layer[25-27].

The analysis method of the number of neurons in the input layer for the evaluation results in the BP neural network subsystem is excellent, good, average, and poor. We have designed four first-level evaluation data volumes and ten secondary indicator layers for the evaluation system. Among them, the secondary evaluation index constitutes ten input vectors of the evaluation neural model. Output port neurons: each evaluation has only one final evaluation result and output hidden layer calculation formula[28]: (21) $\sqrt{n_{j} + n_{0}} + a, 0 ⩽ a ⩽ 10$

n_j represents the count of input neurons, ‘no’ denotes the number of output neurons, ‘a’ remains constant, and the range of values spans from 0 to 10.

Operation of BP neural system hidden layer and sample data normalization preprocessing, combined with the test results, hide and output layers, tensing, and pure line are transfer functions, trail functions are network training functions, learned functions are learning functions, and performance functions are MSE functions.

The weight change of each learning cycle depends on the learning speed. High learning rates can lead to instability or spread of the network; Low learning speed leads to long learning cycles and slow convergence speed, but it can ensure that the amount of errors in the network does not change from error interval to minimum error value. The lower the learning rate, the longer the system lasts. When designing neural networks, network training requires different learning speeds. The secondary distance and the coefficient of decline are observed after each lesson to determine whether the chosen academic achievement is correct. If the error is secondary and decreases rapidly, the learning coefficient is appropriate; If the error is square or fluctuating, the learning coefficient is too high. The more complex the network, the faster the learning speed to support different positional error levels. To determine the adaptive learning rate at different training stages, we adopt the adaptive learning rate. See Figure 7 for the AI entrepreneurship and innovation model[29-31].

4

Analysis of the management and application of the Internet of Things in entrepreneurship education resources

4.1

Data source of model application

Utilizing a provincial university as a case study, the university’s management of innovation and entrepreneurship resources is converted into an Internet of Things framework, and the resultant computational data from innovation education resources of Normalization has been achieved at the university. The distribution of resources for innovation and entrepreneurship education in provincial colleges and universities primarily encompasses these concerns. Initially, the efficiency in utilizing resources is minimal. In that order, the overlaid worth of the Inte The input values for innovation and entrepreneurship education resources in colleges and universities, as per net of Things, stand at 4.08, 4.10, and 4.15 [32]. Additionally, the university’s demographic composition stands at 1:0.80:0.72, in contrast to a mere 1:0.4:0.23. Observations reveal that C8 possesses fewer full-time educators and those with a business background, particularly in creativity and entrepreneurship, with clear distinctions in the input of resources across various colleges and universities. Significant disparities exist in how resources are distributed among colleges and universities. Consider, for instance, the allocation of teachers per student, the fraction of special funds designated for each student, and the area of each student’s internship base for C3 and C9 students are 1:1.2:594.12 and 1:0.7:180.97, respectively. There are large differences in resource allocation [33].

Select the questionnaire data of innovation practice education and entrepreneurial ship quality evaluation of relevant universities as the sample data[34]. 13 data samples were randomly selected through data normalization. The evaluation options are excellent/good/average/poor [35].

4.1.1

Pre-evaluation module

Evaluate the innovation and entrepreneurial skills of the tutors and the student learning competitions. The teacher self-assessment module provides immediate feedback on teaching methods and their effectiveness. The panel comprises economic experts, academics, and government experts and is responsible for evaluating teaching strategies, teaching methods, and their competitiveness. Comprehensive evaluation of teachers’ quality is usually carried out in the middle and before the end of some courses[36].

4.1.2

MATLAB simulation analysis

The new function of the Matlab toolbox is used to build the network. According to the sample data, this function can automatically indicate the number of nerve layers in the starting layer. The number of hidden layers of neurons and the function of the training algorithm must be specified by ourselves. Use the initialize function of the Matlab toolbox. Select the first eight data groups from the specified data samples, form a network on Matlab, and perform predictive tests on the last five data groups to verify the test results.

4.1.3

Analysis of test results

In neural networks, sample training is essential data. Standardized input samples ensure data is within the specified range [0.1], facilitate data processing, and improve network efficiency. The standardized formula of the input sample is as follows: (21) $X = \frac{T - T_{\min}}{T_{\max} - T_{\min}}$

The training results are shown in Figure 8 below

Following three cycles, the error in the system equaled zero. 000135 (<0.001). The discrepancy between the ultimate value of the training sample data and the actual experimental result aligns with the predetermined value. The output value of the system closely mirrors the actual. The significance of data from teaching assessments indicates that the developed BP neural network subsystem more accurately mirrors the needs of the training sample data set. Experimental findings indicate that the neural network evaluation system serves as a model for training evaluation, holding significant theoretical and policy value. Consequently, a neural network-based system for assessing teaching effects is deemed sensible and a Model for assessing the impact of directive teaching. Evaluating the caliber of innovation and entrepreneurial spirit in higher education institutions is a crucial metric for gauging the quality and standard of school teaching.

5

Conclusion

The swift advancement in Internet of Things technology has significantly eased the handling and distribution of resources while introducing fresh avenues for the administration of innovation and entrepreneurial education materials. The research examined the growth traits of present-day local innovation and entrepreneurship training, discovering issues like inadequate resource use and lack of resources. Capable distribution of educational resources. Tracking and analyzing node information for innovation and entrepreneurship educational materials are conducted using the Internet of Things technology. The Internet of Things sensor node’s anti-interference algorithm, educational resource management framework, the AI innovation and entrepreneurship information analysis model, and its advantages. The potential and benefits of employing the AI Internet of Things in innovation and entrepreneurship education have been formulated. Addressing the present challenges in assessing innovation and entrepreneurship education, along with the disproportionate allocation of resources, this document develops an evaluation of innovation and entrepreneurship regarding a framework derived from the artificial intelligence Internet of Things. Ultimately, the study delves into the real-world use of the crafted model for innovation and entrepreneurship education, focusing on resource management within the Internet of Things framework. The following main conclusions can be drawn: 1)

It has constructed the innovation and entrepreneurship education resource management model under the Internet of Things technology, constructed the innovation and entrepreneurship education resource Internet of Things management node, implemented the construction with Web services and SOA, and realized the comprehensive tracking management of data collection, repair application and borrowing management information for innovation and entrepreneurship education resource management.

2)

By improving LANDMARK to assist the actual position of the reference anchor point of the IoT node sensor and then obtain reliable auxiliary positioning RSSI signal value, the mapping reliability is improved, and the IoT node’s ability to resist external environmental interference and positioning accuracy can be effectively enhanced.

3)

Artificial intelligence is introduced into the IoT innovation and entrepreneurship education resources IoT management system, and the artificial intelligence analysis VGG model of IoT nodes is built. The model can analyze IoT management data related to innovation and entrepreneurship education and smartly optimize IoT node management.

4)

The artificial intelligence analysis model of the Internet of Things data for innovation and entrepreneurship education resource management is run through the actual data for simulation, and the simulation prediction results of teaching effect evaluation are generated through the composite mapping of nonlinear action functions. The target values of the model established this time are in good agreement, with the correlation number reaching 0.9094, close to 1. Therefore, the model is objective, scientific, and practical.

Language:: English

Publication timeframe:: 1 times per year
Journal Subjects:: Life Sciences, Life Sciences, other, Mathematics, Applied Mathematics, General Mathematics, Physics, Physics, other

Journal RSS Feed

Analysis of the integration and application of artificial intelligence Internet of Things technology in the construction mode of innovation and entrepreneurship education

Zhongwang Li

Xiaohui Ma

Published Online: Feb 27, 2025

Received: Oct 28, 2024

Accepted: Jan 26, 2025

DOI: https://doi.org/10.2478/amns-2025-0084

KeywordsInternet of Things, artificial intelligence, Anti-interference of IoT nodes, Innovation and entrepreneurship education, BP neural network, Improved LANDMARK location algorithm

© 2025 Zhongwang Li et al., published by Sciendo

This work is licensed under the Creative Commons Attribution 4.0 International License.

Keywords
Internet of Things, artificial intelligence, Anti-interference of IoT nodes, Innovation and entrepreneurship education, BP neural network, Improved LANDMARK location algorithm