Research On The Impact Of Science And Technology Policy Combined With Grey Relation And Deep Learning

Policy is an important tool of modern state governance.As a special kind of policy,science and technology policy plays a role of promoting,supporting,standardizing and guiding the development direction in influencing innovation behavior and achievements.All countries attach great importance to it.At the same time,statistical indicators reflect the social situation in the way of concept and numerical value,which also has high research significance.This paper takes the scientific and technological policy of Hebei Province and the relevant statistical index data of "Hebei Provincial Scientific and Technological Innovation Data Integrated Management System" as the research object and puts forward a classification model of science and technology policy and a method to analyze the impact of science and technology policy on statistical indicators.The main work is as follows:(1)Science and Technology Policy Classification ModelFirstly,the research needs to preprocess the policy text.The word segmentation tool is used for science and technology policies,and Ernie based on Transformer framework is used for vector representation training.Then a feature enhancement classification model based on deep pyramid convolutional neural network is proposed for automatic parsing of policy text.The TF-IDF was integrated into the attention mechanism to amplify the effect of keywords in policy statements on the classification model,and the deep pyramid convolutional neural network was used to extract relevant features of policy statement classification.In order to solve the problem of slow model start up and gradient dispersion,the model introduced residual network to improve the model performance and learn more features in the process of iterative module circulation.In the same environment,the model in this paper is compared with the traditional convolutional neural network,long short-term memory network and deep pyramid convolutional neural network.The experimental results prove that the model proposed in this paper can more accurately complete the task of science and technology policy analysis,and draw the conclusion that deep convolutional neural network is better than traditional cyclic neural network in the simple science and technology policy classification problem.(2)Analysis and Forecast of the Impact of Science and Technology Policy on Statistical Indicators.A method to analyze the influence of science and technology policy on statistical indexes is proposed.On the basis of the completion of the classification of science and technology policies,the science and technology policies of each classification and the statistical indicators of each direction are quantified year by year through the way of proportion.The grey correlation analysis method was used to analyze the correlation degree between each category of the two groups of data,and the policy category was correlated with the statistical indicators through the correlation degree.Then,an index impact prediction model based on BIGRU-CNN is proposed to predict the impact of policy statements on each index.The model uses bidirectional gating cyclic neural network to capture sequence information.By combining the attention mechanism of TF-IDF,the model highlights the keyword information and optimizes the process of feature word extraction.The final representation of science and technology policy statements is obtained by using the local features of CNN network learning policies.The effectiveness of the proposed method and prediction model is verified by comparative experiments and tests.The results show that the proposed method can play a certain auxiliary role in policy analysis.