Historical Inertia: An Ignored but Powerful Baseline for Long Sequence Time-series Forecasting

by   Yue Cui, et al.

Long sequence time-series forecasting (LSTF) has become increasingly popular for its wide range of applications. Though superior models have been proposed to enhance the prediction effectiveness and efficiency, it is reckless to ignore or underestimate one of the most natural and basic temporal properties of time-series, the historical inertia (HI), which refers to the most recent data-points in the input time series. In this paper, we experimentally evaluate the power of historical inertia on four public real-word datasets. The results demonstrate that up to 82 be achieved even by adopting HI directly as output.


page 1

page 2

page 3

page 4


Conditional Time Series Forecasting with Convolutional Neural Networks

We present a method for conditional time series forecasting based on the...

Time Series is a Special Sequence: Forecasting with Sample Convolution and Interaction

Time series is a special type of sequence data, a set of observations co...

From Known to Unknown: Knowledge-guided Transformer for Time-Series Sales Forecasting in Alibaba

Time series forecasting (TSF) is fundamentally required in many real-wor...

Data Smashing 2.0: Sequence Likelihood (SL) Divergence For Fast Time Series Comparison

Recognizing subtle historical patterns is central to modeling and foreca...

An exploratory time series analysis of total deaths per month in Brazil since 2015

In this article, we investigate the historical series of the total numbe...

Construction Cost Index Forecasting: A Multi-feature Fusion Approach

The construction cost index is an important indicator in the constructio...

Model-Attentive Ensemble Learning for Sequence Modeling

Medical time-series datasets have unique characteristics that make predi...