Abstract
As China faces an increasingly competitive employment landscape, the government and educational institutions have pivoted toward encouraging student entrepreneurship. However, while interest is high, success rates remain low, often due to a lack of "psychological construction"—the mental resilience and awareness required to navigate business failure. This paper proposes a hybrid model combining Wavelet Transforms and Neural Networks (WNN) to predict and evaluate the entrepreneurial psychology of college students, specifically focusing on law students. The research identifies four key dimensions: entrepreneurial awareness, volition, ability, and personality. Findings suggest that while academic major has little impact on entrepreneurial readiness, factors like gender and educational level significantly influence a student's "entrepreneurial will" and self-efficacy. The WNN model demonstrates high predictive stability, offering a robust tool for real-time psychological assessment.
What Problem Does This Paper Address?
In China, roughly 70% of students express an interest in starting a business, yet the success rate remains below 10%. The authors argue that the primary bottleneck isn't a lack of policy or funding, but a lack of psychological preparedness. Traditional psychological assessments rely on subjective, manual surveys that lack timeliness and fail to capture the complex, non-linear relationships between various psychological factors. This paper seeks to modernize these evaluations using Artificial Intelligence to provide more accurate, real-time insights into a student's entrepreneurial potential.
Methods: The Wavelet Neural Network (WNN)
The core innovation of this research is the application of a Wavelet Neural Network (WNN). Traditional Backpropagation Neural Networks (BPNN) often suffer from slow convergence and a tendency to get stuck in local minima. By replacing the standard excitation function in the hidden layer with a wavelet function, the model becomes better at processing non-stationary, volatile signals—characteristics common in psychological data.
The wavelet component allows the network to better localize features in both time and frequency domains, leading to higher prediction accuracy and faster learning. The model evaluates four specific dimensions: motivation (Awareness), perseverance (Volition), management skills (Ability), and adventurous spirit (Personality).
Datasets and Reliability
The study utilized a 49-question survey based on the "Entrepreneurship Motivation Structure Measurement Scale." The model was trained and tested using data from law graduates at a specific Chinese institution. Statistical analysis was performed using SPSS 17.0 to validate the reliability and validity of the survey instrument itself.
As shown in the data analysis, the Cronbach’s α reached 0.82, indicating that the four dimensions of the evaluation system are statistically sound and reliable for predictive modeling.
Key Insights and Results
- The "Master’s Peak" and "Doctoral Dip": One of the most striking findings was that students with a master’s degree exhibited the strongest entrepreneurial will, while doctoral students had the lowest. This suggests that over-specialization or the high opportunity cost of a PhD might diminish the desire to take entrepreneurial risks.
- Gendered Self-Efficacy: While men and women scored similarly across most dimensions, men showed significantly higher self-efficacy (confidence in one's own ability to succeed) within their psychological capital ($P < 0.05$).
- Major Neutrality: Interestingly, the student's specific major did not have a statistically significant impact on their overall entrepreneurial psychological scores. This suggests that "entrepreneurial spirit" is a broader psychological trait rather than a byproduct of specific disciplinary training.
- Model Superiority: The WNN model proved highly stable, with the difference between predicted values and actual psychological scores staying within 0.3 points on a 5-point scale. The authors conclude that the wavelet-based approach avoids the "blindness" of traditional structural design in neural networks, making it more suitable for the "vague" and "volatile" nature of human psychology.
Limitations and Future Work
The authors acknowledge that the study relied on a relatively small sample from a single university. Future research needs larger, more diverse datasets to generalize these findings across different regions. Additionally, the paper suggests that follow-up research should track students over time to see how their entrepreneurial psychology evolves as they move from the classroom into the actual business world. By moving psychological evaluation from subjective reports to algorithmic prediction, this research paves the way for universities to provide targeted, AI-driven support to students.
Summary updated