Algorithms for big data mining of hub patent transactions based on decision trees

¹ Expert and Analytical Center, 33, Talalikhina str., Moscow, 109316, Russia
² Federal Research Center “Computer Science and Control” of Russian Academy of Sciences, 40, Vavilov Street, Moscow, 119333, Russia
³ Central Economics and Mathematics Institute of Russian Academy of Sciences, 47, Nakhimovsky Prospekt, Moscow, 117418, Russia
⁴ National University of Science & Technology MISiS, 4, Leninsky Prospect, Moscow, 119049, Russia
⁵ Marine Hydrophysical Institute, Russian Academy of Sciences, 2, Kapitanskaya str., Sevastopol, 299011, Russia

^* Corresponding author: kartsan2003@mail.ru

Published online: 17 February 2025

Abstract

Dysfunctions of the patent supply and demand market have a negative impact on the sustainability of the national innovation system, which stimulates the growth of prices for knowledge-intensive products. It is necessary to establish a relationship between fiscal decisions regarding patent transactions and the prospects for the development of commercialization of the results of intellectual activity. One of the most promising methods for improving the accuracy of system analysis of big and semi-structured patent transaction data is the use of decision trees. Existing methods based on the error backpropagation method are quite slow, and their accelerated versions lose in training accuracy. To effectively solve the problem of forecasting the cost of hub patent transactions, parametric algorithms have been developed based on response bias and with the additional use of predicative structures of the model of successive geometric transformations. The optimal number of decision tree predicates has been established taking into account computational efforts and the accuracy of forecasting the cost of hub patent transactions. Based on evolutionary computing, the optimal values of the parameters of algorithms for big data mining of hub patent transactions have been established.