Empirical Techniques for Enhanced Predictive Modeling: Beyond Traditional ARMA

Author(s): Shenggang Li

Originally published on Towards AI.

This member-only story is on us. Upgrade to access all of Medium.

A Non-Parametric Approach for Robust Forecasting and Data Analysis Across Domains

The ARMA model is a popular choice for time series forecasting because it captures how data points are related over time — like how today’s data depends on yesterday’s. But ARMA assumes that the residuals follow a specific (usually normal) distribution. In real-world data, this assumption often doesn’t hold up. Outliers, sudden shifts, and unusual patterns can mess with the model, making forecasts less accurate or unstable.

From my research, I believe that empirical techniques offer a solution. They’re flexible, non-parametric, and adapt directly to the data without needing strict distribution assumptions. Instead of forcing data into a set framework, they use the actual observed values to build the model, making it effective at handling outliers and complex patterns that traditional models might miss.

Combining empirical transformation and likelihood estimation with ARMA leads to a more reliable forecasting model. ARMA captures time-based relationships, while empirical likelihood helps manage irregularities. Instead of assuming a specific residual distribution, empirical likelihood lets the model adapt to real-world data, improving forecast accuracy.

Let’s dive into the world of empirical distribution and… Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI