Methods for Patent Landscape Modelling and Predictive Analysis

Ilya Kozlov, Proletarsky Andrey, Artyom M. Panfilkin, ELENA V. SMIRNOVA, Berezkin V. Dmitry, Robin Clark

Keywords: clustering, machine learning, patent analysis, trend forecasting

TL;DR: The article presents new methods for constructing a patent landscape model, a method for assessing development directions in technological areas, and a subsystem for visualizing the results of analysis and forecasting of technological trends.

Abstract: The article is dedicated to the problem of forecasting changes in the patent landscape based on the analysis of multimodal data. This article examines three main approaches to the analysis of patent information: 1) a clustering-based approach; 2) a resource-based approach; 3) a machine learning approach. Each approach is considered in terms of its advantages, flaws and potential for use in predicting the emergence of new technologies. The article proposes new methods for constructing a patent landscape and predictive models, as well as a method for assessing development directions in technological areas. The article also considers the problem of visualizing the results of analysis and forecasting of technological trends. To solve this problem, the authors propose a visualization subsystem that allows graphical demonstration of the changes in the patent landscape over time. The article presents the results of experiments demonstrating the system's ability to make correct predictions of technological trends.

Primary Area: applications to computer vision, audio, language, and other modalities

Submission Number: 14179