Introduction

With the development of science and technology, machine learning (ML) has become a critical technology for processing tasks across various industries. As a kind of ML, deep learning (DL) technology has significantly contributed to the development of all industries. Meanwhile, entrepreneurship plays a vital role in industry growth, and research into its dynamics has a considerable impact on shaping the future of entrepreneurship. The investigation into the path of college students’ entrepreneurship education provides a key reference for improving the quality of entrepreneurship education and the sustainability of entrepreneurial ventures. Furthermore, examining the management strategies of start-up enterprises offers insights for guiding their future development and serves as a feedback mechanism for refining college students’ entrepreneurship education. Although the current research on college students’ entrepreneurship education and social entrepreneurship through DL and artificial intelligence (AI) technology is not mature, existing studies offer valuable references for advancing this important field.

Although there is no perfect specific index for college students’ entrepreneurship education, many studies provide a reference for the path of college students’ entrepreneurship education. Lin and Ma (2018) argued that entrepreneurship education was a crucial component of quality education in colleges, which was of great significance in improving students’ entrepreneurial abilities. Innovation and entrepreneurship education are gradually becoming integral to talent training in colleges. To improve the actual effect of entrepreneurship education, based on the understanding and analysis of the necessity of an entrepreneurship education ecosystem, a complete “trinity” entrepreneurship educational ecosystem should be constructed; The government, colleges, and society should make joint efforts to enhance the effect of entrepreneurship education. Innovation and entrepreneurship are key drivers of economic development and social progress1. Gao (2020) highlighted that the education department should reflect on the current situation of China’s college entrepreneurship practice system, such as insufficient teachers, students’ incorrect professional concepts, biased theories, and others. Moreover, they proposed several countermeasures, including strengthening curricula, building innovation and entrepreneurship platforms, expanding practical training bases, and supporting entrepreneurship projects2. Zeng (2019) observed that the current research and development of science and technology were accelerated, leading to the influx of all kinds of advanced science and technology into all walks of life, especially the “Internet +” technology. This technology could be introduced into innovation and entrepreneurship education for college students to provide better education and guidance3. Jin et al. (2019) pointed out that with the continuous evolution of the times, the ability of innovation and entrepreneurship became the key to reflecting the comprehensive national strength. In response to China’s call to build an “innovative country”, colleges should cultivate students’ personalized thinking, methods, problem-solving skills, and creativity, laying a solid foundation for their future entrepreneurial and innovative capabilities. These studies provided critical insights into the direction of college students’ entrepreneurship education4. Parallel to research in entrepreneurship education, many studies have examined the management strategies of start-up enterprises. Shepherd et al. (2020) identified several key factors that influenced the success of start-up enterprises, including chief founder, founding team, social relations, cognition, emergency organization, new risk strategy, organizational emergence, legitimacy, founder exit, and entrepreneurial environment5. Du and Kim (2021) pointed out that new enterprises could face diverse competitions and external challenges when seeking high performance. To navigate this, they required various market and non-market growth strategies tailored to their complex environment6. Rocha et al. (2019) emphasized that instead of randomly adopting various strategies, managers should consider their impact on performance. Hence, they argued that strategic choice was endogenous in the performance equation. Although various scholars have made increasing efforts to solve endogenous bias, previous attempts have almost completely focused on unilateral and discrete (binary) organizational decision-making. However, managers often face multiple, simultaneous, and interdependent decisions in reality, which may include a continuous selection set. These choices may further involve a two-way process between managers and others (such as employees, strategic partners, customers, or investors), and their choices and preferences can significantly affect the final decision-making, thus influencing the survival of the enterprise7. In short, the current research into college students’ entrepreneurship education and start-up enterprise management strategies offers valuable insights for the development of entrepreneurship projects. However, the study of the broader social entrepreneurship landscape remains underdeveloped, and there is no specific method to solve the problems faced by the current social entrepreneurship situation. The integration of AI and DL technologies has great advantages in data processing. It is an innovative measure to analyze the factors of college students’ entrepreneurship education and start-up enterprise management strategies.

To sum up, first, the path of college students’ entrepreneurship education is studied. The factors influencing entrepreneurship education are analyzed through DL and AI methods, and the comprehensive factors are summarized. Next, based on feedback regarding these influencing factors, the path of entrepreneurship education is analyzed in detail. Subsequently, the start-up enterprise management strategies are studied. The current situation and sources of risk within start-up enterprises through the deep neural network (DNN) and AI algorithm. Through the analysis and summary of the final influencing factors, the risks are predicted and reasonable strategies are put forward. In the research content, AI technology and DL algorithms are integrated, innovating the specific methods of college entrepreneurship teaching, which plays an important role in reforming traditional teaching. This study provides valuable guidance for improving college students’ entrepreneurship education and supports the enhancement of start-up enterprise management strategies.

Research background

Causal attribution theory

College students’ entrepreneurship education paths and start-up enterprise management strategies are explored based on the causal attribution theory. First, the entrepreneurial psychology of college students is investigated through a questionnaire survey and literature review. Moreover, the collected data are analyzed and summarized through DL and AI algorithms to identify a better path for college students’ entrepreneurship education. Then, a start-up enterprise management platform is constructed through AI and DNNs, and the management strategy of start-up enterprises is examined based on causal attribution theory.

The process begins by designing a questionnaire based on established theories and previous research findings, followed by a review and refinement of the questionnaire. Next, respondents or groups are identified, and questionnaires are distributed. Then, the questionnaires are collected, and the obtained data are processed and analyzed8. The literature review process involves five factors: topic proposal, research design, literature review, data collection, and data processing9. Ultimately, the data from the literature review and questionnaire survey are summarized, analyzed, and processed, and the results are discussed.

It is essential to understand the causal attribution theory in analyzing the management strategies of start-up enterprises. Causal attribution theory refers to the psychological process through which individuals infer the causes of outcomes based on internal information such as reasoning, perception, and thought. This process involves analyzing the causes of behaviors or events. The theory is mainly divided into attribution cognitive process theory and attribution effect theory. The former explains the attribution process, focusing on how individuals interpret their own and others’ behavior. The latter refers to connecting attribution with its impact on people’s behavior and psychology, providing insights into its broader effects10. This study applies these theories to explore start-up enterprise management strategies and college students’ entrepreneurship education paths.

Research elements of entrepreneurship education

The first research method employed in this study is a literature review. This involves collecting targeted literature, analyzing and identifying relevant sources, and selecting the appropriate literature for analysis. The review aims to assess the current state of research, identify key problems, research routes, and outcomes, and establish a theoretical foundation for the study. The second method is a questionnaire survey, which gathers insights into contemporary college students’ entrepreneurial ideas. Through this survey, the current state of entrepreneurship education is assessed. Then, the results of the literature research and questionnaire are integrated. The aspects that contemporary entrepreneurship education should focus on are studied. The data processing method for analyzing the collected insights involves examining the interactive effects of different factors11. Figure 1 demonstrates the influence of various factors on the path of college students’ entrepreneurship education.

Fig. 1
Fig. 1
Full size image

Effect diagram of the impact of entrepreneurship education path.

Figure 1 illustrates the interaction mechanism of multiple factors influencing students’ entrepreneurial psychological environment. As shown in the figure, the external environment, family background, personal factors, and school environment collectively form a complex network that affects students’ psychological environment. Among these, personal factors include age, gender, and entrepreneurial motivation, which exert their influence at the individual trait level. Age is associated with experience and risk tolerance, while gender differences may lead to varying entrepreneurial preferences and decision-making styles. Entrepreneurial motivation directly drives the initiation and persistence of entrepreneurial behavior. Regarding the school environment, the level of education reflects the extent to which teaching resources, faculty strength, and academic atmosphere support the development of students’ entrepreneurial literacy. Different academic disciplines, with their distinct knowledge systems and practical orientations, shape students’ unique thinking patterns and skill sets, thus influencing their entrepreneurial perspectives and path choices12. In this influence chain, students’ psychological environment occupies a central role, integrating the effects of external factors and feeding them back into the outcomes of entrepreneurship education. These outcomes, as key outputs, further influence the final choice of the entrepreneurship education path13. Given this complex relationship, the selection and management of entrepreneurial paths for college students must be analyzed from multiple dimensions, carefully weighing the weight of various influencing factors. A deep exploration of their intrinsic logic and potential effects is essential, providing a solid theoretical and practical foundation for determining the most suitable entrepreneurship education path. Thus, it strengthens the foundation for students’ entrepreneurial practices14. Table 1 exhibits the various conditions of the college students studied.

Table 1 Basic information of respondents.
Full size table

In Table 1, the scientific rigor and representativeness of sample selection are critical in this study. To comprehensively analyze the various factors affecting student entrepreneurship and explore suitable entrepreneurship education paths, this study adopts a multi-channel, hierarchical sampling method, selecting 300 students with entrepreneurial experience from a wide range of institutions. In terms of age, the sample spans from lower undergraduate years to graduate students, based on educational stages and entrepreneurial development patterns, to capture the dynamic evolution of entrepreneurial intentions and abilities with age. Regarding gender, the principle of gender balance is strictly followed to ensure that both male and female perspectives are adequately represented. This enables a deep exploration of the potential impact of gender differences on entrepreneurial decision-making, resource acquisition, and risk preferences. Considering family background, the study comprehensively considers factors such as urban-rural differences, economic class diversity, and the presence or absence of a family entrepreneurship tradition. It precisely analyzes how family values, resource support, and social networks shape the initiation and development of entrepreneurship. In terms of academic discipline, the study includes a broad spectrum of majors, including science and engineering, business management, and humanities and social sciences. It reveals the unique contributions and constraints of different knowledge systems, thinking models, and practical orientations in the entrepreneurial process. Regarding school type, both key and general institutions are considered, comparing the differential impacts of educational resources, academic atmosphere, and competitive pressure on the entrepreneurship education ecosystem. This ensures that the 300 samples comprehensively represent the entrepreneurial ecosystem of college students. Thus, it provides foundational data for subsequent research and enables a deep analysis of the factors influencing entrepreneurship education and the optimization paths for such education.

In the research process for start-up enterprise management strategies, constructing a network model based on DNNs and AI is a core element. To ensure the reliability and universality of the research conclusions, the selection of start-up enterprise samples follows strict standards and scientific procedures. The sample selection begins with an extensive screening of commercial databases and industry recommendations, encompassing start-up enterprises established within a specific time range (typically 1–5 years) across various industry sectors. This stage focuses on factors such as enterprise size, development stage, and market potential to ensure sample diversity and representativeness, capturing the commonalities and unique features of start-up enterprise management strategies. In terms of size, the sample includes enterprises ranging from micro-teams to 100 employees, analyzing the differences in management strategies under different human resource structures. In terms of the development stage, the study targets enterprises at key stages, from product development and market testing to early profitability, to uncover the dynamic evolution of management strategies. Regarding market potential, the study evaluates enterprises based on industry growth rates, market share expansion expectations, and technological innovation assessments. It includes high-potential and stable growth enterprises, to explore the role of external environments in shaping management strategies15. The sample selection also emphasizes the observability of management strategies and data availability, prioritizing enterprises with relatively complete structures, clear decision-making processes, and standardized data records, ensuring the depth and quality of the data. For determining the success of management strategies, a multidimensional evaluation framework is constructed. Short-term performance indicators include market share growth rate, year-on-year revenue increase, and optimization of customer acquisition costs, which measure the ability to build market competitive advantages. From a long-term perspective, brand value enhancement, continuous innovation capabilities, and customer loyalty maintenance are evaluated16. In addition, internal organizational coherence, employee satisfaction, and talent retention rates are key indicators that reflect the effectiveness of management strategies in fostering team cohesion and vitality. This comprehensive evaluation of the management strategy’s effectiveness in balancing multiple objectives lays a strong foundation for subsequent DNN and AI model analyses. It can accurately identify factors that affect start-up enterprise management strategies, and conduct effective risk assessment. Meanwhile, it can provide targeted guidance for optimizing and upgrading start-up enterprise management strategies, and enhance enterprise survival and development capabilities. The abstract concept of the DL algorithm is displayed in Fig. 2.

Fig. 2
Fig. 2
Full size image

Operation flow chart of DNNS.

Figure 2 shows multiple layers of DL, the most complex of which is the hidden layer. DL covers multiple hidden layers, each containing multiple factors. While the factors within a single layer are not interconnected, the layers themselves are connected. Data processing occurs sequentially, with each layer processing the data independently. Factors within the same layer do not influence each other. Next, the AI encoder used here is based on an unsupervised learning and training method, which automatically performs calculation through data input and subsequently outputs the data results17. Using this method to analyze the influencing factors of college students’ entrepreneurial path and the management strategy of start-up enterprises can generate accurate results. Figure 3 presents the basic workflow of the AI encoder.

Fig. 3
Fig. 3
Full size image

Workflow of AI automatic encoder.

Figure 3 reveals that the automatic encoder includes two layers: the encoder and the decoder. The error ultimately comes from the comprehensive analysis of the two. After undergoing unsupervised learning, the AI encoder operates without manual intervention. Once the data is provided, the system automatically analyzes it and produces error results18. This study offers vital technical support for the data analysis process using AI. Moreover, data processing through AI ensures higher accuracy for the research outcomes.

Research method

This study uses the causal attribution theory to investigate college students’ entrepreneurship education paths and start-up enterprise management strategies. First, the path of college students’ entrepreneurship education is studied through the interaction effect. The study includes two variables, school education and external environmental impact. The calculation is based on these two comprehensive factors, as listed in Eq. (1):

$$\:M{S}_{t}=a+{b}_{1}SM{M}_{t}+{b}_{2}M{Q}_{t}+{b}_{3}{V}_{t}+{c}_{t}+{d}_{t}$$
(1)

\(\:M{S}_{t}\) represents the entrepreneurial motivation of college students; \(\:SM{M}_{t}\) is the level of school education; \(\:SM{M}_{t}*M{Q}_{t}\) refers to the interaction between school education and external influence. \(\:{V}_{t}\) stands for the key factors affecting college students, encompassing family background, school, major, age, and gender. \(\:a\) means the psychological factors of students; \(\:b\) represents the basic state of students; \(\:{c}_{t}\) reveals the imperceptible potential influencing factors; \(\:{d}_{t}\) denotes other small probability random influencing factors. The adjustment equation, Eq. (2), is also constructed to make the model more accurate and applicable:

$$\:M{S}_{t}=a+{b}_{1}SM{M}_{t}+{b}_{2}M{Q}_{t}+{b}_{3}{V}_{t}+{b}_{4}SM{M}_{t}*M{Q}_{t}+{c}_{t}+{d}_{t}{V}_{t}$$
(2)

A comprehensive calculation of the impact of school education and the external environment is added. The research on college students’ entrepreneurial path starts with multiple factors. Based on the causal attribution theory, the factors affecting college students’ entrepreneurship education paths are comprehensively collected. The data are processed based on linear regression algorithm through the DNN and AI model, and then the data processing results are analyzed. Finally, the research findings are summarized to determine the comprehensive factors affecting the path of college students’ entrepreneurship education.

The deep neural calculation method is to correct the weight and deviation of each neuron. The specific principle is to judge the error function of the whole neural network, and correct it if the function decreases. The specific calculation equation reads:

$$\:{X}_{k+1}={X}_{k}-{a}_{k}{b}_{k}$$
(3)

\(\:{X}_{k}\) represents the deviation value and weight of the network; \(\:{X}_{k+1}\) refers to the calculated deviation value and weight; \(\:{a}_{k}\) stands for the learning speed of the neural network; \(\:{b}_{k}\) denotes the gradient of the error function. The neural network inputs data during operation and the calculation method of each network layer is as follows:

The input of the neuron \(\:i\) in the first layer of the hidden layer can be written as Eq. (4):

$$\:{{u}^{I}}_{i}={\sum\:}_{g=1}^{G}{W}_{gi}{X}_{kg}$$
(4)

W represents the connection weight between all levels of neural networks. The output of the neuron \(\:i\) in the first layer of the hidden layer is:

$$\:{{\upsilon\:}^{I}}_{i}=n\left({\sum\:}_{g=1}^{G}{W}_{gi}{X}_{kg}\right)$$
(5)

The output of the neuron \(\:t\) in the second layer of the hidden layer reads:

$$\:{{\upsilon\:}^{I}}_{t}=n\left({{\sum\:}_{i=1}^{I}{W}_{it}{V}_{i}}^{I}\right)$$
(6)

The input of the neuron \(\:d\) in the output layer is:

$$\:{{u}^{D}}_{d}={\sum\:}_{t=1}^{I}{W}_{td}{{V}_{t}}^{T}$$
(7)

The expression of the output of the neuron \(\:d\) in the output layer reads:

$$\:{m}_{kd}={{v}_{d}}^{D}=n\left({\sum\:}_{t=1}^{T}{W}_{td}{{V}_{t}}^{T}\right)$$
(8)

The output error of the neuron \(\:d\) in the output layer is as follows:

$$\:{e}_{kd}\left(h\right)={d}_{kd}\left(h\right)-{y}_{kd}\left(h\right)$$
(9)

The synthesis of output errors of all neurons in the output layer is shown in Eq. (10):

$$\:E\left(h\right)=\frac{1}{2}{\sum\:}_{d=1}^{D}{{e}^{2}}_{kd}\left(h\right)$$
(10)

The above discussion highlights that the DNN calculation involves transmitting signals layer by layer while continuously correcting errors. After processing through all the layers, the total output error is ultimately aggregated at the output layer19.

The data needs to be verified in data analysis. The verification equation of normal distribution is as follows:

$$\:{F{{max}|}_{v}{\left(x\right)}_{o}x)}_{max}$$
(11)

\(\:{S}_{v}\) and \(\:{E}_{o}\) represent two assumptions; \(\:{F}_{max}\) denotes the highest point of data distribution. Mann-Whitney U test algorithm is used for the nonparametric test.

$$\:{V}_{\alpha\:}={x}_{\alpha\:}{x}_{\beta\:}+\frac{{x}_{\alpha\:}\left({x}_{\alpha\:}+1\right)}{2}-{Y}_{\alpha\:}$$
(12)
$$\:{V}_{\beta\:}={x}_{\alpha\:}{x}_{\beta\:}+\frac{{x}_{\beta\:}\left({x}_{\beta\:}+1\right)}{2}-{Y}_{\beta\:}$$
(13)

\(\:{V}_{\alpha\:}\) and \(\:{V}_{\beta\:}\) refer to the maximum values of different data; \(\:{x}_{\alpha\:}\) and \(\:{x}_{\beta\:}\) represent the individual data of diverse combinations, respectively. \(\:{Y}_{\alpha\:}\) and \(\:{Y}_{\beta\:}\) are the average of each data group, respectively. The comprehensive factors are analyzed through data calculation, and the final results are studied.

This study employs a comprehensive range of scientific research methods to lay the foundation for exploring college students’ entrepreneurship education paths and start-up enterprise management strategies. During the data collection phase, a questionnaire was designed and widely distributed, targeting a sample of 300 students with entrepreneurial experience and diverse backgrounds. The sample encompasses a broad spectrum of differences across age, gender, family background, major, and school type, enabling a deep exploration of factors influencing entrepreneurship. Concurrently, a systematic literature review is conducted, following a five-step process to extract key elements of entrepreneurship education and core principles of enterprise management from the literature, thus providing a theoretical foundation for the study.

In the data analysis phase, linear regression analysis is employed as the primary research method to examine the interactive effects of school education, external environment, and individual student factors. This method precisely quantifies the weight of each factor’s influence on entrepreneurial motivation, allowing for the accurate identification of key driving factors. DL algorithms, leveraging the advantages of multilayer architectures, including input, hidden, and output layers, explore the complex relationships of factors within the hidden layers. Guided by the neuron error function, the signals are transmitted and corrected progressively through the layers, enabling efficient processing of high-dimensional complex data. The models for entrepreneurship education paths and enterprise management strategies are dynamically optimized in real-time, capturing subtle changes in variables and long-term trends, thus enhancing the models’ predictive and explanatory power. For model validation, normal distribution validation equations and Mann-Whitney U tests are used to rigorously check the normality of data distributions and the significance of inter-group differences. This ensures the robustness, reliability, and generalizability of the research results. These methods provide strong support for the practical innovation of college students’ entrepreneurship education and the scientific optimization of start-up enterprise management strategies, offering deep empowerment for developing entrepreneurial ecosystems and innovation.

Modeling based on causal attribution theory

This study employs the causal attribution theory to examine start-up enterprise management strategies and college students’ entrepreneurship education paths. First, the factors influencing college students’ entrepreneurship education paths are modeled using the DNN algorithm and AI processing. Multiple factors affect the entrepreneurship education path of college students, making it crucial to establish a research and analysis model through the interaction effect20. Figure 4 reveals that the results are influenced by both the external environment and school education. The relative impact of each factor must be determined through data analysis.

Fig. 4
Fig. 4
Full size image

Comprehensive influence diagram of College Students’ entrepreneurship education path.

Figure 4 illustrates that based on the causal attribution theory, school education factors and external environmental factors collaboratively influence students’ entrepreneurial mindset, entrepreneurial outcomes, and the path selection of entrepreneurship education. Students’ entrepreneurial mindset impacts entrepreneurial outcomes, which in turn determines their path choice of entrepreneurship education. The interplay of multiple factors creates a complex impact, necessitating comprehensive research from various perspectives21.

The research and modeling of management strategies for start-up enterprises aim to collect and analyze the influencing factors on the management strategies of various start-up enterprises, thereby establishing models for influencing factors and risk management strategies22. Figure 5 shows the risks of start-up enterprises from different aspects.

Fig. 5
Fig. 5
Full size image

Risk management model of start-up enterprises.

Figure 5 shows that the risks of start-up enterprises are pervasive, with risks in different areas directly impacting the survival of these enterprises. Based on causal attribution theory, management strategy risk, system risk, and environmental risk are the main risk sources. These risks include secondary risks, with management strategy risks being particularly significant. These risks encompass decision-making errors, cultural differences, internal staff resistance, and employee turnover. These stimulus risks may lead to major risks at any time, and eventually bring serious risks to start-up enterprises. The study of start-up management strategies involves a detailed, step-by-step analysis of risk points in the management strategy, based on secondary risks. After a thorough evaluation, the impact of management strategy risks on start-up enterprises is predicted, and practical recommendations are made to mitigate risks, helping enterprises identify better management strategies and ensuring steady progress in the future. A comprehensive study is made according to the basic situation of students’ mentalities, such as age, gender, family background, major, and school. These factors are analyzed to evaluate their impact on college students’ entrepreneurship education and start-up enterprise management.

Research results

The path of college students’ entrepreneurship education based on causal attribution theory

Various factors affecting college students’ entrepreneurship are comprehensively analyzed through student surveys under different conditions. The final results are obtained by analyzing external factors and their factors, which are presented in tables following data analysis. Table 2 reveals the factors that influence college students’ entrepreneurship.

Table 2 Influence coefficients of various influencing factors on college students’ entrepreneurship.
Full size table

Table 2 lists the influencing factors, including age, gender, family background, major, and school. By analyzing the number of entrepreneurial failures in each category and making a comprehensive comparison, it is concluded that each influencing factor has varying degrees of influence. An analysis chart is made for a comprehensive comparison of the data in the table, highlighting the most crucial factors. The degree of influence of each factor is suggested in Fig. 6.

Fig. 6
Fig. 6
Full size image

Analysis of influencing factors.

Figure 6 shows significant variations in the influence of different factors. Among them, the smallest comprehensive influence coefficient is the age of college students, and the direct and indirect influence coefficients are below 0.20. The largest direct influence coefficient is the students’ major, up to 1.86, followed by family background (1.24) and school (1.29), respectively. The largest indirect influence coefficient is attributed to gender, with a value of 1.75, followed by family background, major, and school, which are 1.15, 1.16, and 0.96, respectively. A comparison of direct and indirect influencing factors with the comprehensive influencing factors is presented in Figs. 7 and 8. The comparison results can more intuitively show the more important factors affecting college students’ entrepreneurship.

Fig. 7
Fig. 7
Full size image

Comparison of direct influencing factors and comprehensive influencing factors.

Figure 7 displays that the age of students has the least influence on the effectiveness of entrepreneurship education, with a direct influence coefficient of around 0.2 and a comprehensive influence coefficient close to 0.1. The most influential factor is the student’s major, which has a direct influence coefficient of about 1.8 and a comprehensive influence coefficient of approximately 3.0. The teaching effects of different majors are very different. This indicates that the teaching effectiveness varies significantly across different majors.

Fig. 8
Fig. 8
Full size image

Comparison of indirect influencing factors and comprehensive influencing factors.

In Fig. 8, the comprehensive comparison shows that the lowest influence coefficient on college students’ entrepreneurial effect is still their age, with a comprehensive coefficient of 0.30. The biggest influence coefficient is associated with the major, and a comprehensive influence coefficient is 3.02, significantly surpassing the influence coefficient of other factors. Next, gender, family background, and school greatly impact college students’ comprehensive impact coefficients, at 2.29, 2.36, and 2.25, respectively.

In-depth analysis of management strategies of start-up enterprises based on attribution theory

The risk occurrence coefficients of start-up enterprises are analyzed based on causal attribution theory, and the data are processed by the DNN. Table 3 shows the occurrence coefficients of various risks faced by start-up enterprises and the corresponding weights of various influencing factors.

Table 3 Various risk occurrence coefficients of start-up enterprises.
Full size table

Table 3 suggests that the coefficients of various influencing factors differ, and the influencing factors of the risk of start-up enterprises show a certain bias. Among them, the lowest and highest risk occurrence coefficients are external environmental and management strategy risks. The weight calculation results show that management strategy risk and institutional risk exert the greatest impact on start-up enterprises, and the calculated weights are 0.35 and 0.30, respectively. Therefore, research on the influencing factors of start-up enterprises should focus on these two aspects. Analyzing the management strategy risk of start-up enterprises is essential for improving their management strategies. Figure 9 demonstrates the influence coefficients and weights of different factors in the start-up enterprise management strategies.

Fig. 9
Fig. 9
Full size image

Influencing factors of management strategy risk.

In Fig. 9, the management strategy of the start-up enterprise is analyzed. The results show that there are several influencing factors, with the most influential factor being poor decision-making by managers, which has a coefficient of 9, followed by employee turnover with a coefficient of 8. The corresponding influence weights of these two factors are also the highest, at 0.175 and 0.138, respectively. The lowest influence weight is the resistance of internal employees, and the weight value is 0.024.

Result analysis

Based on the comprehensive and in-depth analysis presented above, this study provides clear directions for college students’ entrepreneurship education and start-up enterprise management strategies. In the domain of entrepreneurship education, the study reveals that various factors such as age, gender, major, family background, and school environment have different impacts on entrepreneurship. Among these, the influence of the academic major is particularly prominent. This finding offers crucial guidance for schools in developing entrepreneurship education programs. Schools should widely promote and strengthen entrepreneurship education for all students, optimizing course offerings and teaching models according to the specific effects of these factors. This approach can help eliminate educational imbalances caused by major-related differences and ensure the fair and efficient advancement of entrepreneurship education across different student groups. Additionally, the deep integration of DL and AI technologies in the selection of entrepreneurship education paths, with their powerful data analysis capabilities, can detect potential issues in the educational process. Moreover, it can precisely pinpoint teaching gaps and student needs, and generate targeted solutions to improve the quality and adaptability of education.

In the dimension of start-up enterprise management strategies, the study identifies managerial decision-making errors and employee turnover as the core risk factors. At the inception of a startup, careful selection of leaders with decision-making and leadership abilities is essential. It is necessary to establish scientific decision mechanisms and supervision systems to reduce the risks associated with decision-making errors from the outset. Furthermore, enterprises should strengthen employee welfare systems, provide comprehensive care for employees, and actively engage in morale-building activities to enhance employee sense of belonging and loyalty, effectively curbing turnover and resistance behaviors. Moreover, the full potential of DL and AI algorithms should be leveraged to continuously monitor the dynamics of management risks, using big data for deep mining and intelligent analysis. This enables the rapid and accurate identification of potential risks, allowing for timely adjustments and optimization of management strategies to ensure stable operations.

Compared to other advanced studies, this study offers significant advantages23,24. Some related studies focus solely on individual factors or specific types of risks, lacking a systematic and comprehensive perspective. This study constructs a multifactorial analytical framework that comprehensively covers the key aspects of entrepreneurship education and enterprise management, quantifying the weight of factor influences and risk coefficients. This provides more precise and actionable strategies for practice. Additionally, the innovative integration of cutting-edge technologies breaks through the limitations of traditional qualitative analyses, enabling a qualitative leap in the depth and breadth of the research-driven by data. This contributes to the innovation of college students’ entrepreneurship education and the optimization of start-up enterprise management strategies, injecting strong momentum into the development of a healthy and sustainable entrepreneurial ecosystem.

Conclusion

Based on causal attribution theory, this study explores college students’ entrepreneurship education paths and start-up enterprise management strategies. By leveraging DNNs and AI to collect and analyze existing data, a comprehensive examination of college students’ status is conducted by combining literature reviews and questionnaire surveys. Meanwhile, AI and DL technologies are used to process and analyze the data in detail.

The study finds that the factors influencing college students’ entrepreneurship education include the internal educational environment, the external environment, and the students’ characteristics. Within the internal educational environment, the differences in professional programs and teaching models across different disciplines result in a significant impact on the effectiveness of entrepreneurship education. In the external environment, the societal environment has a more prominent effect on the effectiveness of entrepreneurship education compared to the family environment. Regarding students’ characteristics, gender remarkably influences entrepreneurial outcomes, while age has a relatively weaker impact. Furthermore, entrepreneurial outcomes reflect the results of entrepreneurship education. For start-up enterprise management strategies, the influence coefficient of managerial decision errors is as high as 9, while employee turnover has a coefficient of 8. Their corresponding impact weights are 0.175 and 0.138, respectively, making them the most significant factors among all variables. In contrast, the influence of internal employee resistance has the lowest weight, at only 0.024. Although DL and AI algorithms play a key role in this study, there are limitations. While the sample selection is representative, it mainly originates from specific regions and types of institutions, which limits the generalizability of the conclusions. The data collection channels focus primarily on questionnaires and public data, potentially overlooking key sources of information. While the research method integrates multiple technologies, qualitative methods such as in-depth interviews and case studies are absent, limiting the exploration of deeper causal mechanisms.

Future research could expand the sample scope to include diverse regions, institutions, and backgrounds to enhance the generalizability of the conclusions. It also enriches the data collection approach, incorporating multi-source data to uncover potential factors. By integrating qualitative and quantitative methods, the research framework on college students’ entrepreneurship education and start-up enterprise management strategies can be improved, providing precise guidance for entrepreneurial practice.