A Novel Evaluation and Prediction Model of Innovation Efficiency Based on GCA-GDEA Index Screening and AdaBoost Integration of Machine Learning

Manufacturing is the foundation of a country. Since the 21st century, China's manufacturing industry has entered a new stage of rapid growth and faces higher development requirements. Therefore, to better evaluate manufacturing enterprises' innovation efficiency, this paper innovatively constructs a systematic enterprise performance evaluation and classification model. Firstly, a preliminary efficiency assessment of the enterprise based on the generalized data envelopment analysis method(GDEA) was to obtain the relative efficiency value. Subsequently, grey relational analysis was used to quantify the degree of correlation between each output indicator and the efficiency value, screen out the key influencing factors, and achieve feature selection. On this basis, a secondary GDEA analysis was implemented to optimize the efficiency assessment results and enhance the ability to identify sample heterogeneity. Next, the efficiency evaluation results are labeled and divided to construct a binary classification dataset. The K-nearest neighbor, support vector machine, and logistic regression model are used for training, respectively, and the parameters are optimized through grid search. Finally, the AdaBoost ensemble learning method is utilized to conduct a weighted fusion of multiple base classifiers, construct a strong learner, and accurately identify high-efficiency enterprises. The results show that: (1) The efficiency evaluation platform is based on the grey correlation degree, and the AdaBoost integrated model has better evaluation and prediction capabilities than the traditional GDEA model. (2) Classify and predict the future development of the research enterprises and conduct specific strategic analysis for enterprises of different classifications.