Scholarlink

Time series are usually affected by missing data, that is there are a few time series samples whose values are not known. This paper presents the Reiterated Deep Gaussian Process (RDGP) algorithm that performs the prediction of time series with missing data. RDGP algorithm employs reiteratively the Correlation Dimension Estimation to fix the model order and a Deep Gaussian Process to approximate the skeleton of time series. RDGP algorithm was experimentally validated on two time series with missing data, measuring the concentration of Ozone in two European sites, showing a small average percentage prediction error for the time series on the test set.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Time Series Prediction with Missing Data by an Iterated Deep Gaussian Process

  • 章节

  • 发表日期：

    2025年

查看全文
  • Francesco Camastra,
  • Angelo Casolaro,
  • Gennaro Iannuzzo

摘要

Time series are usually affected by missing data, that is there are a few time series samples whose values are not known. This paper presents the Reiterated Deep Gaussian Process (RDGP) algorithm that performs the prediction of time series with missing data. RDGP algorithm employs reiteratively the Correlation Dimension Estimation to fix the model order and a Deep Gaussian Process to approximate the skeleton of time series. RDGP algorithm was experimentally validated on two time series with missing data, measuring the concentration of Ozone in two European sites, showing a small average percentage prediction error for the time series on the test set.