Predicting the innovation activity of chemical firms using an ensemble of decision trees

A number of studies are concerned with the analysis of predicting innovation activity, because companies' innovation activity is one of the fundamental determinants for their competitiveness. However, most studies use a linear (logistic) regression model for their analysis. This, however, is not able to take into account all the recursive terms concerning a company's innovation activity. Therefore, in the report we demonstrate the use of ensembles of decision trees to model the intrinsic nonlinear characteristics of the innovation process. We apply this method for predicting innovation activity to chemical companies. We show that internal knowledge spillovers were the most important determinant for the chemical Arms' innovation activity during the monitored period. Furthermore, R&D intensity, collaboration on innovation and firm size were also important determinants.

[1]  Z. Griliches,et al.  Do Subsidies to Commercial R&D Reduce Market Failures? Microeconomic Evaluation Studies , 1999 .

[2]  John C. Platt,et al.  Fast training of support vector machines using sequential minimal optimization, advances in kernel methods , 1999 .

[3]  Alberto López,et al.  Understanding co-operative innovative activity: evidence from four European countries , 2009 .

[4]  Petr Hájek,et al.  Competitive advantage analysis: a novel method for industrial clusters identification , 2012 .

[5]  M. Nieto,et al.  The importance of diverse collaborative networks for the novelty of product innovation , 2007 .

[6]  M. Frenz,et al.  The impact on innovation performance of different sources of knowledge: Evidence from the UK Community Innovation Survey , 2009 .

[7]  Leo Breiman,et al.  Bagging Predictors , 1996, Machine Learning.

[8]  K. Arrow.  Economic Welfare and the Allocation of Resources for Invention , 1962 .

[9]  Thomas G. Dietterich.  Multiple Classifier Systems , 2000, Lecture Notes in Computer Science.

[10]  Tai-Yue Wang,et al.  Forecasting innovation performance via neural networks—a case of Taiwanese manufacturing industry , 2006 .

[11]  S. van Buuren.  Multiple imputation of discrete and continuous data by fully conditional specification , 2007, Statistical methods in medical research.

[12]  Hans Schaffers,et al.  Smart Cities and the Future Internet: Towards Cooperation Frameworks for Open Innovation , 2011, Future Internet Assembly.

[13]  Jan Stejskal,et al.  Descriptive Analysis of the Regional Innovation System - Novel Method for Public Administration Authorities , 2013 .

[14]  Petr Hájek,et al.  Predicting Firms' Credit Ratings Using Ensembles of Artificial Immune Systems and Machine Learning - An Over-Sampling Approach , 2014, AIAI.

[15]  Yoav Freund,et al.  A decision-theoretic generalization of on-line learning and an application to boosting , 1995, EuroCOLT.

[16]  Petr Hájek,et al.  Visualising components of regional innovation systems using self-organizing maps—Evidence from European regions , 2014 .

[17]  Hanna Hottenrott,et al.  (International) R&D Collaboration and SMEs: The Effectiveness of Targeted Public R&D Support Schemes , 2012 .

[18]  Volkswirtschaftliche Diskussionsreihe,et al.  R & D cooperation and innovation activities of firms : Evidence for the German manufacturing industry , 2002 .

[19]  L. Rubalcaba,et al.  Knowledge for innovation in Europe: The role of external knowledge on firms' cooperation strategies , 2013 .