Author(s): Shenggang Li

Originally published on Towards AI.

Adapting Diverse Models to Transform Transactional Data for Predictive Accuracy
Photo by Elena Mozhvilo on Unsplash

Time series forecasting is fundamental to predictive modeling, drawing significant interest from data scientists. Its applications range from supply chain demand prediction to financial market forecasting. Unlike traditional models, time series forecasting models can be unstable due to the inherent variability of time.

In a recent post, Time Series Forecasting: A Comparative Analysis of SARIMAX, RNN, LSTM, Prophet, and Transformer, I conducted a comparative analysis of various time series forecasting methods. While I should have included a machine-learning approach, the post became too lengthy. Therefore, I am presenting this paper today to address that gap.

In this post, I’ll keep exploring time-series models, how they work, and why I proposed a generalized machine-learning framework for forecasting transactional data using feature engineering. While applying machine learning to time series forecasting is not new, the generalized framework I propose aims to address practical time series forecasting challenges across various scenarios.

Models for Out of Sample vs. Out of Time

Forecasting and prediction are often used interchangeably, but they have distinct meanings in time series analysis. Forecasting involves predicting future data points by analyzing past and present data over time and projecting future values based on historical trends. This is why models… 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