A nonlinear multivariate grey Bernoulli model for predicting innovation performance in high-tech industries

Abstract

As China modernizes its industries, predicting innovation performance in high-tech industries is essential for crafting innovation-driven strategies. However, the system output of high-tech industries is influenced by multiple input factors with interaction effects, often exhibiting non-linearity and uncertainty. To address this, a novel nonlinear multivariate grey Bernoulli model considering interaction effects, IENGBM(1,N), has been developed. This model identifies interaction effects between factors and the nonlinear features of the system, more flexibly capturing the fluctuating and nonlinear trend of innovation performance. This paper conducts a quantitative selection of modeling data instead of the former, which is a more rigorous way. Moreover, a comprehensive framework integrating particle swarm optimization (PSO), Monte Carlo simulation, and probability density analysis (PDA) is designed to enhance and validate the model's predictive accuracy. The model's solution is optimized, effectively eliminating jump errors. Experimental results show that the IENGBM(1,N) model with strong robustness outperforms five compare models in two cases, with fitting MAPE of 1.97 % and 2.97 %, and testing MAPE of 1.61 % and 1.44 %, respectively, and this verifies the absence of overfitting in the model. Lastly, the proposed model has been utilized to forecast the innovation performance of high-tech industries and their sub-industries in the feature, providing a practical and effective prediction tool.