A Common Trap and Countermeasures in Application of Machine Learning in
Machine learning (ML) algorithms have gained more and more successful
applications in the field of energy prediction. However, current
conventional application process of ML algorithms lacks screening of
dominant factors and model validation, resulting in weakening the
predictive ability of the model. In this work, systematic and robust
predictive models are provided to address this issue. Based on 147 sets
of data, various methods were used to predict. The results show that in
the process of learning curve analysis, the eight models can get
satisfactory results, but there are three models with overfitting or
underfitting. Besides, through analyze the model by the single-factor
control variables method, two additional defective models were found.
Therefore, current conventional ML modeling methods are not reliable.
This paper addresses the main reasons for the poor performance of some
predictive models built by ML and provides guidelines on how to build
robust predictive models.