Visual Time Series Forecasting: An Image-driven Approach
This approach offers a novel way to handle time-series data for researchers in forecasting, though it is incremental in applying vision techniques to this domain.
The paper tackles time-series forecasting by treating it as a computer vision task, using images to predict distributions, and finds that the method outperforms baselines like ARIMA in image-based metrics, especially for cyclic data.
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.