Design and Implementation of University Students Scientific Research Ability Evaluation System Based on Neural Network

Table 3

Evaluation errors of different prediction models.

Data sources	Error metrics	Traditional network	Network optimized by gradient descent	Network optimized by quasi-Newton method	Network optimized by GOA	Our model
1	MSE	0.3257	0.3428	0.2786	0.3287	0.2238
	RMSE	0.6258	0.5276	0.5376	0.5837	0.4834
	MAE	0.4853	0.4685	0.3825	0.4127	0.3658
2	MSE	0.4615	0.3275	0.2685	0.4835	0.2385
	RMSE	0.6813	0.5684	0.4752	0.6952	0.4537
	MAE	0.5368	0.4376	0.4576	0.5276	0.3864