Research On Evaluation Method Of Children's Logical Thinking Ability Based On Deep Neural Network

Logical thinking ability is one of the most important abilities in children's daily learning and life.In this Internet era,the cultivation of logical thinking ability has been paid more and more attention in various industries of the whole society,such as education industry,medical industry,etc.The cultivation of logical thinking ability is highly valued in children's learning stage,and it will also play a great role in children's later growth process.Western countries which attach importance to independent and independent education have a relatively high degree of popularization of logical thinking ability.Logical thinking ability plays an important role in every stage of children's growth and development,but so far,China has not set up courses similar to the improvement of logical ability.At present,the efficiency and accuracy of some popular evaluation methods are very low.How to improve the accuracy is by no means easy.The evaluation results will also be affected by many factors such as the evaluation methods and the emotions of the evaluators.At the same time,the current popular shallow learning model can not deeply explore the chaotic relationship between various features,especially when this kind of model infers high-dimensional sparse features,its prediction results are not satisfactory.However,based on the emergence of deep learning,the influence of many subjective factors will be appropriately reduced.Therefore,label inference based on algorithm model becomes the focus of children's logical thinking ability evaluation.This paper analyzes the reason why shallow learning has no obvious effect in predicting high-dimensional sparse features,and finds a method combining deep neural network and gradient lifting decision tree to predict the strength of children's logical thinking ability.In the experiment,the deep neural network is used as the feature extractor,and the classifier selects the gradient lifting decision tree with the advantages of ensemble learning.Experiments show that the deep neural network model is more effective than shallow learning in the face of high-dimensional sparse features,and it is easier to get relatively accurate experimental results.At the same time,the deep BP neural network can improve the inference results of the shallow learning model.Therefore,the experimental results of this model are the most intuitive and accurate by combining the deep neural network with the gradient lifting decision tree.The experimental content of this paper provides a new research direction for the estimation of children's logical thinking ability.In addition,it is necessary to introduce the ID feature in the modeling process.