Research on Talent Training Effect and Mode Innovation under the Background of Artificial Intelligence

Research areas:

• Volume 14,Issue 1, July 2025

Year:

2025

Type of Publication:

Article

Keywords:

AI Talent Training, XGBoost Model, SHAP Value Analysis, Education Resource Allocation Optimization

Authors:

Yibei Lin; Mingzhe Li

Journal:

IJAIM

Volume:

14

Number:

1

Pages:

1-10

Month:

July

ISSN:

2320-5121

Abstract:

This study aims to identify key factors influencing the effectiveness of interdisciplinary AI talent training in China and to provide data-driven strategies for educational improvement. To achieve this, a predictive model of talent training outcomes was developed using XGBoost, with interpretability enhanced through SHAP (SHapley Additive exPlanations) analysis. Data were collected via stratified sampling from 386 students, followed by preprocessing and feature selection using Pearson correlation and mutual information regression. The XGBoost model demonstrated superior performance, reducing prediction error by 22.7% after hyperparameter optimization compared to baseline models. Results reveal that institutional tier is the most influential factor: graduates from 985 universities earn 18.7% more than those from 211 universities and 31.2% more than non-211/985 graduates. Furthermore, a "golden triangle" of core course satisfaction, internship duration, and mathematical foundation was found to increase expected salaries by 28.3%. Based on these findings, the study proposes actionable strategies, including dynamic curriculum updates, strengthened industry-academia collaboration, and targeted resource allocation, to enhance AI talent development in higher education.

Full text: IJAIM_688_FINAL.pdf