Visual Time Series Forecasting: An Image-driven Approach

TL;DR

This paper proposes a novel approach to time-series forecasting by transforming data into images and using computer vision techniques, which improves distribution prediction and outperforms traditional methods on cyclic datasets.

Contribution

It introduces an image-driven method for time-series forecasting that captures data as images and trains models to predict future images, enhancing distribution forecasting capabilities.

Findings

Effective for cyclic data

Outperforms ARIMA and other baselines on image-based metrics

Less effective for irregular data like stock prices

Abstract

In this work, we address time-series forecasting as a computer vision task. We capture input data as an image and train a model to produce the subsequent image. This approach results in predicting distributions as opposed to pointwise values. To assess the robustness and quality of our approach, we examine various datasets and multiple evaluation metrics. Our experiments show that our forecasting tool is effective for cyclic data but somewhat less for irregular data such as stock prices. Importantly, when using image-based evaluation metrics, we find our method to outperform various baselines, including ARIMA, and a numerical variation of our deep learning approach.