Design And Implementation Of Patent Map Software System

As a new innovation system,regional innovation has been paid more and more attention by the government,and the biggest problem regional innovation facingis the introduction of experts.Patent document,as an effective carrier of information technology and expert information,can provide effective talents information for regional innovation.Patent map is an important patent analysis tool.It can induce,classify and analyze the disordered patent information and show the results in the form of charts.Especially in the patent map,the inventor analysis chart and the patentee analysis chart can analyze the distribution of talent information well in a technical field,thus providing an effective talent introduction program for the government.There are many researches on patent maps both at home and abroad,and the content of the study has been gradually transferred from the traditional patent map research to the study of patent content itself.However,most of these researches are based on the traditional vector space model,and the data mining methods used are relatively simple,and the mining of patent content itself is not deep enough.In addition,there are few researches on the analysis of the inventor and the patentee.The researches are still in the scope of the traditional patent map,and can not analyze the expert information from different technical perspectives.Therefore,it is necessary to make a deep study of the analysis chart of the inventor and the patentee.The main contributions of this thesis are as follows:(1)A feature engineering method based on N-gram,TF-IDF and Word2Vec is proposed in the thesis.The N-gram model can extract the keywords of the patent abstract and title,and then combine TF-IDF method with Word2Vec to calculate the patent text vector,finally the patent text vector matrix which is the input for next step of clustering was obtained.(2)After obtaining the patent text vector matrix,the K-means clustering algorithm is used to cluster the patent text,and comparing with the traditional feature engineering methods,such as:LDA,LSA and VSM model.Experiments show that the feature engineering method used in this thesis can achive the best result of clustering.(3)Based on the clustering results,conducting the study of the inventor and the patentee map and the patent scatter map,fillay drawing up the patent map by Echarts.This thesis designs and implements an analysis of patent map software system on the patent inventor and patentee.The experimental results show that this system can make inventor and the patentee map better,then providing an intuitive analysis.The feature engineering method used in the system can improve the accuracy of clustering with certain application value.