(paper) Forecasting with Time Series Imaging

Forecasting with Time Series Imaging (2019, 21)

Contents

1. Abstract
2. Introduction
3. Time Series Imaging & Feature Extraction
4. Overall Algorithm

0. Abstract

• 1) transform TS to “Recurrence Plots”
• 2) extract features from Recurrence Plots
• 3) extracted features are used for “Forecast Model Averaging”

1. Introduction

• TS clustering

• TS classification

• Anomaly detection

\(\rightarrow\) quantification of “SIMILARITY among TS data”

1) most of existing approaches depend on manual choice

2) current literature on feature-based forecasting focuses on “global” features, not “local”

\(\rightarrow\) automated feature extraction becomes vital

Sections

• Section 2 : feature extraction from image

• Section 3 : how to assign weights to a group of forecasting methods ( using image features )

• Section 4 : application

2. Time Series Imaging & Feature Extraction

(1) Recurrence Plots

3. Overall Algorithm