An adaptive heterogeneous ensemble learning method for multi-dimensional company performance decision-making

Evaluating company performance is central to strategic decision-making and sustainable development. Given the limited development of comprehensive investigations into the impact of multidimensional variables on company performance within existing research, we pioneer the use of machine learning methods to explore this issue. To accurately predict company performance in terms of operating income, net profit, and total assets, we propose a novel adaptive heterogeneous ensemble learning (AHEL) method that adaptively outputs the heterogeneous ensemble learning model with the best predictive performance for each predicted aspect. Experimental results on real-life data from 740 Chinese listed companies over the period 2010–2020 demonstrate that AHEL outperforms several state-of-the-art machine learning methods for multi-dimensional company performance prediction and leads to better organizational decisions. We also examine the relative importance of the features of the predicted aspect and interpret the correlations between the important feature values and the prediction results of AHEL. The findings reveal that the features ‘social responsibility’, ‘shareholder responsibility’ and ‘R&D expenditure’ all positively impact the predicted results, ‘government subsidy’ has a threshold effect on the predicted results, and ‘digital transformation’ and ‘innovation’ have a mixed impact on the predicted results. These prescriptive insights enhance researchers’ understanding of multi-dimensional company performance prediction that benefits their future work.