Influence of artificial intelligence on higher education reform and talent cultivation in the digital intelligence era

Datasets assortment

The info studied in this paper primarily comes from open academic platforms and databases, overlaying three varieties of datasets. Firstly, the dataset of academic sources is collected by means of MOOC platform and the nationwide platform of higher education knowledge education. After cleansing, greater than 11,000 information of on-line programs, educating supplies and educating movies are saved. Secondly, the educating mannequin datasets of totally different majors come from China Higher Education Pupil Data Community, which covers the educating plans and curriculum construction knowledge of greater than 200 majors in schools and universities throughout the nation. After standardization, they’re mixed for mannequin evaluation³⁵. As well as, as a result of the restricted entry to the attribute knowledge of college students’ conduct educating, 8,000 college students’ studying conduct knowledge are captured and simulated by means of the educating information made public by native ministries of education, and mixed with info comparable to research habits, attendance price and grades, a multi-dimensional scholar studying dataset is shaped. After knowledge cleansing, 37,892 legitimate information are lastly retained, and the variables with correlation coefficient larger than 0.3 are screened by correlation evaluation in the characteristic extraction stage. After knowledge preprocessing, 28,6062 knowledge are retained for mannequin coaching and testing.

The info info of coaching and testing is proven in Desk 1:

Experimental atmosphere

The coaching and testing of SEOM mannequin relies on the core objective of mannequin design, and its efficiency is comprehensively and objectively examined from a number of ranges and angles. Firstly, in phrases of the accuracy and generalization capacity of the mannequin, cross-validation (Okay = 10) and set-aside validation are adopted to check its capacity to keep up excessive prediction accuracy and robust generalization below totally different knowledge distributions³⁶. In the meantime, primarily based on Root Imply Sq. Error (RMSE) and Imply Absolute Error (MAE), the error of the mannequin is quantified to confirm its capacity to cope with multi-dimensional academic sources and complicated educating fashions. Secondly, in the side of studying path optimization, by calculating Cmin/df, Goodness of Match Index (GFI), Adjusted Goodness of Match Index (AGFI), Comparative Match Index (CFI), Root Imply Sq. Residual (RMR) and Approximate Root Imply Sq. Error (RMSEA), and analyze the advice impact of the mannequin on personalised studying paths in totally different main classes (engineering, science, medication, legislation, education, administration, literature, economics, artwork and agriculture) in widespread educating scenes^37,38,39. As well as, by setting an irregular community atmosphere, Shapley worth is used to research the contribution and delay sensitivity of useful resource allocation, and the clever impact of useful resource allocation in complicated atmosphere is examined^40,41.

Amongst them, the analysis standards of every index in the evaluation of studying path optimization impact are proven in Desk 2:

Parameters setting

On this paper, the Graph Consideration Community (GAT) is chosen as the core sort of GNN, and the options of neighboring nodes are given totally different weights by means of the consideration mechanism to seize the uneven dependence between nodes extra precisely. GAT is very appropriate for personalised studying path planning and academic useful resource optimization as a result of of its robust expressive capacity and adaptability to complicated relationships between nodes. GAT is used to mannequin the multi-level interplay between college students and data factors, dynamically regulate the weights of adjoining nodes by means of consideration mechanism, and generate a characteristic vector illustration that’s extra appropriate for the precise studying scene. Its structure consists of three layers of consideration, and every layer adopts a multi-head consideration mechanism to enhance the robustness and world nature of characteristic seize. In the utility atmosphere, GAT first receives the high-quality options screened by RFA as enter and constructs the preliminary graph construction. Then, by means of iterative characteristic updating and aggregation, the dependency map of studying path is steadily optimized, and lastly personalised advice outcomes are generated. The primary cause for selecting GAT is that it may well successfully cut back the over-smoothing drawback in contrast with the conventional GNN, particularly in the dynamic education scene, and it may well deal with the complicated interplay between college students’ studying conduct and data factors extra precisely by means of the propagation of node traits and the weighted replace of edges. SEOM hyperparameters and coaching strategies are organized as proven in Desk 3:

Related parameters of widespread educating scenes in the evaluation and verification of mannequin studying path optimization impact are proven in Desk 4:

The related parameters of sudden irregular community atmosphere in the evaluation and verification of mannequin academic useful resource allocation impact are proven in Desk 5:

In fact, in order to raised spotlight the benefits of SEOM in this paper, referring to the literature evaluate in the earlier paper, 5 baseline fashions which are most related to this paper are chosen, in contrast and verified below the similar knowledge normal (10,000 knowledge in the dataset are randomly chosen). The chosen baseline fashions are sorted out, and the outcomes are proven in Desk 6:

The software program and {hardware} atmosphere parameters in the complete coaching and testing of the mannequin are proven in Desk 7:

Efficiency analysis

Evaluation of accuracy and generalization capacity of the mannequin

The evaluation outcomes of SEOM mannequin accuracy and generalization capacity are proven in Fig. 2:

In Fig. 2, SEOM mannequin exhibits excessive accuracy and good generalization capacity in cross-validation (Okay = 10) and set-aside validation. The RMSE worth of cross-validation fluctuates between 0.2 and 0.5, and the MAE worth is between 0.1 and 0.5, which exhibits the stability of the mannequin in coping with multi-dimensional academic sources and complicated educating fashions. In the meantime, the accuracy of the mannequin stays at 85-97%, indicating its reliability in optimizing academic sources and recommending studying paths. The comparability between RMSE and MAE is barely higher, which retains fluctuating between 0.2 and 0.6, however inside the accuracy vary of 80-90%, indicating that the mannequin nonetheless has robust adaptability and huge utility potential.

Impact evaluation of mannequin studying path optimization

The evaluation end result of SEOM mannequin studying path optimization impact is proven in Fig. 3:

Determine 3 exhibits that the Cmin/df of SEOM mannequin is between 1.0 and 2.5 in totally different main classes and educating scenes, indicating that the mannequin has a perfect becoming diploma in every scene. The indexes of GFI and AGFI are above 0.85, and CFI is near 0.95, which exhibits that the mannequin has excessive accuracy and rationality in capturing the dependence of data factors in totally different educating modes. In the meantime, RMR and RMSEA values are decrease than 0.05, which exhibits that the residual of the mannequin is small and the mannequin has robust adaptability to the precise educating scene. Particularly in the personalised educating scene, the advice impact of the mannequin on the studying path is considerably improved, which additional verifies SEOM’s clever optimization capacity in the complicated academic atmosphere.

Impact evaluation of mannequin education useful resource allocation

The evaluation outcomes of academic useful resource allocation effectivity of SEOM mannequin are proven in Fig. 4:

In Fig. 4, the Shapley worth ranges from 0.1 to 0.4 in the irregular community atmosphere, indicating that there are vital variations in the contribution of totally different community environments to useful resource allocation. Amongst them, bandwidth bottleneck and community delay have nice affect on the effectivity of useful resource allocation, and the delay sensitivity is excessive, as much as 0.8, which exhibits that community delay has a powerful interference impact on the allocation of educating sources. The effectivity of useful resource allocation fluctuates between 60% and 95%, which relies upon on the stability of the community atmosphere. The evaluation outcomes present that SEOM can intelligently optimize the allocation of academic sources below complicated community situations, particularly in the atmosphere of excessive delay and bandwidth bottleneck, and the mannequin exhibits robust adaptability and useful resource allocation capacity to make sure the efficient completion of educating duties.

Comparative evaluation of mannequin with different baseline fashions

The comparative evaluation outcomes of SEOM mannequin and different baseline fashions are proven in Fig. 5:

As Fig. 5 exhibits, SEOM mannequin has apparent benefits in many key indicators, particularly when coping with multi-dimensional academic sources optimization and personalised studying path advice, and its efficiency is superior to the conventional baseline mannequin. In phrases of mannequin accuracy, the accuracy of SEOM is at all times above 95% when the knowledge scale is steadily elevated to 10,000, which is considerably higher than the secure values of GCN and SVM fashions of 88% and 83%, displaying glorious generalization capacity and robustness. At the similar time, in the modeling of complicated dependencies, SEOM successfully optimizes the distribution of characteristic weights by advantage of the consideration mechanism of GATT. The RMR worth is lowered to 0.035, which is considerably decrease than the common worth of 0.052 in different baseline fashions, reflecting that the becoming error of the mannequin to multidimensional knowledge is considerably lowered. As well as, in the optimization of useful resource allocation effectivity and studying path, SEOM maintains the allocation effectivity of greater than 90% in the excessive knowledge density atmosphere by means of the dynamic weighted adjustment of Shapley worth. Moreover, its useful resource adaptation capacity is extra outstanding than that of KNN mannequin, the place the effectivity drops to 75%. This exhibits that SEOM optimizes the complicated dependence amongst data factors in the personalised studying scene, and successfully realizes the clever allocation of educating sources in the complicated atmosphere. This absolutely displays its huge utility potential in the clever reform of higher education.

Dialogue

SEOM mannequin combines RF, AdaBoost and GNN to construct an clever optimization system for complicated academic sources and personalised studying paths. RF improves the accuracy of multi-dimensional academic sources processing by means of adaptive enhancement mechanism, and GNN enhances the accuracy of the mannequin in studying path prediction by setting up the dependency map between college students and data factors. The mannequin verification outcomes present that the becoming indexes comparable to Cmin/df, GFI, AGFI and CFI are extremely correct, and the indexes of RMR and RMSEA are in an affordable vary, which signifies that SEOM has excessive becoming capacity and small error. Particularly in the irregular community atmosphere, SEOM exhibits glorious useful resource allocation optimization capacity by means of Shapley worth and delay sensitivity evaluation. As well as, SEOM can intelligently advocate the optimum studying path in accordance with college students’ studying habits and data factors in personalised studying scenes, and notice the seamless docking of personalised educating by means of real-time knowledge replace. In the side of academic useful resource administration, SEOM intelligently allocates and transmits all types of sources by integrating with the administration platform to make sure the environment friendly allocation of sources and the easy progress of educating duties.

From the optimization level of view, SEOM successfully alleviates the bottleneck drawback of characteristic extraction in conventional education mannequin when coping with high-dimensional and complicated knowledge by means of the deep mixture of improved RF algorithm and adaptive enhancement mechanism. RF algorithm improves the robustness of characteristic choice by means of multi-tree construction, and adaptive enhancement mechanism additional strengthens the complete efficiency of weak classifiers, enabling SEOM to precisely seize key options in academic sources and optimize useful resource allocation methods. On this foundation, GNN, by advantage of its modeling capacity of dynamic dependency, not solely comprehensively depicts the complicated interplay relationship, but in addition dynamically adjusts the weight of nodes by setting up the map construction between college students and data factors. It successfully copes with the drawback of nonlinear characteristic distribution and dynamic change in studying path advice. The verification outcomes additional illustrate this level. SEOM considerably reduces the error price (RMSE worth is lowered to 0.2–0.5) in high-noise knowledge atmosphere. Its environment friendly becoming capacity is verified by Cmin/df and GFI, particularly in complicated community situations, which may precisely optimize the useful resource allocation effectivity by means of Shapley worth and delay sensitivity evaluation. In contrast with the conventional mannequin, the structure design of SEOM has proven excellent benefits in clever useful resource allocation, personalised studying path advice and adaptability of educating situations, thus selling the complete realization of clever optimization of higher education.