Application of improved random forest algorithm to enterprise architecture performance evaluation

To achieve performance evaluation of YG Power Enterprise, a performance evaluation system has been developed. Adopting a random forest algorithm improved based on genetic algorithm for performance evaluation, using entropy method and data envelopment analysis Malmquist index model to evaluate the total factor productivity index of enterprise performance indicators. The results show that the genetic algorithm has a significantly better parameter optimization effect on the random forest model than the grid search algorithm. The accuracy of genetic algorithm is as high as 85.84%, which is 2.15% higher than grid search algorithm. In the comparison of feature selection performance, the algorithm used in the study has a curve area of up to 0.8025. Applying this algorithm to the actual performance evaluation of YG power enterprises, it can be concluded that the total factor productivity of the data branch is as high as 1.072, followed by the industrial branch with a value of 1.613. The total factor productivity of the energy subsidiary is only 0.925, which is mainly constrained by technological progress. The combination of improved random forest algorithm and data envelopment analysis Malmquist index model can effectively achieve comprehensive evaluation of YG power enterprise performance, providing technical support for strategic adjustment of the enterprise.