一类多区间预测的深度学习模型及实验分析

A deep learning model for multi-interval forecasting and experimental analysis

文章基于高质量区间原则,考虑不同置信水平下区间之间的关系,将单区间损失函数推广到适用于多区间的损失函数,并利用公开的家电能耗数据进行深度学习和对比实验分析.结果表明,多区间损失函数不仅能够有效解决单区间的越界交叉问题,而且能够高效合理地量化预测的不确定性.

Based on the high-quality prediction intervals principle,this paper considers the relationship between intervals that are at different confidence levels,generalizes the single-interval loss function to a multi-interval loss function.Using the appliances energy prediction data set,this paper presents a deep learning experiment as well as a comparative experimental analysis.The obtained results show that the multi-interval loss function can effectively solve the crossover problem and quantify the uncertainty of prediction more efficiently and reasonably.