An Intelligent Evaluation Algorithm of Practical Innovation Ability for Students

Abstract

The existing evaluation method indicators are not specific and the index weights are highly subjective. This paper selects the evaluation indicators to estimate the practical innovation ability of graduate students and undergraduates, and proposes an evaluation algorithm on the basis of deep belief network (DBN), and an improved algorithm based on practical innovation ability model of graduate students. Since the evaluation indicators and data distribution of undergraduate students are very similar to those of graduate students, the improved algorithm adopts the parameter based transfer learning method. The weight of the same characteristics of undergraduates and graduate students is directly multiplied by the difference factor as the initial weight of the undergraduate fine-tuning. The weight of disparate characteristics for undergraduates and graduates needs to be fine-tuned and re-trained. Experiment results show the improved algorithm has wider application ranges and higher accuracy rate, overcomes the problem of strong subjectivity about index weights, and it is beneficial to promote reform of talent training and the overall improvement of talent training quality. The comprehensive evaluation algorithms on the basis of fuzzy mathematics, the evaluation algorithms of probabilistic neural networks, the general deep learning evaluation algorithms, and the presented algorithm are compared to verify the effectiveness of the proposed evaluation algorithm.