Identifying and Leveraging Leading Indicators for Time Series Forecasting
Introduction
On a daily basis, companies are faced with decisions relating to manoeuvring financial markets, optimising their supply chain operations, or formulating strategies to maintain a competitive edge amidst evolving trends. Nevertheless, achieving high accuracy results in forecasting can prove elusive when time series models fail to account for interconnected events or other time series that exert influence on the subject of prediction.
In this article, we will explore the concept of leading indicators, how to identify them, and how to leverage them to improve time series forecasting. We will dive into the practical implementation using Python and real-world data sourced from the Federal Reserve Economic Data.
What are Leading Indicators?
Leading indicators are datasets that help in forecasting future trends or activities. An everyday example of a leading indicator is the sudden appearance of cloud cover, which could signal the likelihood of a thunderstorm occurring within the next hour.
How to Identify & Leverage Leading Indicators
- Domain Knowledge: Like all Data Science projects, we start by understanding the domain we will be operating within. A key aspect of this process is to identify variables or factors that are likely to influence the time series we want to forecast. In our practical implementation, we will be forecasting Beer, Wine, and Liquor salesᵈ¹ so factors like Consumer Price Index⁵
