The Patent Technology Opportunity Recognition Based On The Outlier Detection

Nowadays, the external environment of enterprise is complex, the development of science & technology improves rapidly, the competition between the enterprises is also growing. The competition between the enterprises in the final analysis is the competition of technology innovation.The one who takes the initiative on technology innovation, will get a place in specific market activities. For that reason, the identification of technology opportunity has become particularly important. Therefore, the significance of this study is to build the recognition model of technology opportunity based on anomaly detection so as to help enterprises identify technology opportunities and gain the initiative in technological innovation activities.As the core carrier of technology, patent text has inherent advantages such as easy access, complete information and so on. With patent text as the research object, this paper puts forward the abnormal meanings in patent text, and tests the patent text by using our technology opportunity recognition model to achieve the purpose of technology opportunity identification. The main content of the paper includes three parts.First, it is the similarity calculation of patent text. In this part, the vector space model is used to represent patent text in the form of vector, then the similarity between patents is computed based on the cosine theorem, and finally, the generated patent similarity matrix goes through multidimensional analysis and visualization.Second, it is the construction of recognition model of technology opportunity based on anomaly detection. After the dimension reduction and visualization based on difference distance between patents, four types of anomaly detection algorithm are used for anomaly detection on the patent data set. The abnormal patent points under each anomaly detection algorithm are screened into the abnormal patent set. Finally, by virtue of experts’ wisdom, we use have a selection and discrimination of abnormal patent sets to recognize technology opportunities.Third, it is the empirical analysis. In order to verify the validity of the method of the paper and to make predictions for the particular technology, we select the study object of patent data in the field of laser lithography. First our recognition model of technology opportunity based on anomaly detection is verified, which shows the validity of the method in this paper. Finally, identification and prediction of technology opportunity are carried out based on recent patent data in the field of laser lithography and good experiment results have been achieved.