article · Journal of Forecasting
The k-nearest neighbours algorithm is a well-known method for classification and regression that can also be applied to time series forecasting. However, choosing the appropriate number of neighbours and selecting suitable features remains difficult. To address this, two new forecasting approaches have been developed: Classical Parameters Tuning in Weighted Nearest Neighbors and Fast Parameters Tuning in Weighted Nearest Neighbors. The classical approach evaluates the most recent data subsequence against every past subsequence of the same length. The fast approach shrinks the search space, reducing the grid size and cutting computational time. Both techniques use a cross-validation approach designed for weighted nearest neighbours to tune parameters. The models were tested against standard methods including Seasonal Autoregressive Integrated Moving Average, Holt Winters, and Exponential Smoothing State Space Models, using real datasets on United States retail and food service sales alongside United Kingdom milk production.
Accurate forecasting helps organisations plan inventory, anticipate demand, and manage operations. While k-nearest neighbours is a simple and widely used machine learning technique, fine-tuning its parameters for time-dependent data is often computationally heavy. Providing faster, reliable ways to calibrate these models makes data-driven predictions more practical across diverse sectors such as retail, food services, and agriculture.
The work represents applied and tested algorithmic research evaluated on real-world industry data, specifically retail and food sales alongside milk production. Organisations and software developers working on operational forecasting or supply chain planning could implement these methods to generate predictions more quickly. Further software engineering and integration into existing business intelligence pipelines would be necessary to reach commercial deployment as a software product.
AI-generated from the published abstract. Always read the original work before citing.
Abstract The k ‐nearest neighbors algorithm is one of the prominent techniques used in classification and regression. Despite its simplicity, the k ‐nearest neighbors has been successfully applied in time series forecasting. However, the selection of the number of neighbors and feature selection is a daunting task. In this paper, we introduce two methodologies for forecasting time series that we refer to as Classical Parameters Tuning in Weighted Nearest Neighbors and Fast Parameters Tuning in Weighted Nearest Neighbors. The first approach uses classical parameters tuning that compares the most recent subsequence with every possible subsequence from the past of the same length. The second approach reduces the neighbors' search set, which leads to significantly reduced grid size and hence a lower computational time. To tune the models' parameters, both methods implement an approach inspired by cross‐validation for weighted nearest neighbors. We evaluate the forecasting performance and accuracy of our models. Then, we compare them to other approaches, especially, Seasonal Autoregressive Integrated Moving Average, Holt Winters, and Exponential Smoothing State Space Model. Real data examples on retail and food services sales in the United States and milk production in the United Kingdom are analyzed to demonstrate the application and efficiency of the proposed approaches.
This page summarises published work. The authoritative version sits with the publisher.
DOI: 10.1002/for.3093
Is something wrong with this record? Report it or request removal.
Discussion
Have you built on this work, tried to replicate it, or seen it applied in practice? Share what you know. Verified researchers and MARATTO™ domain experts can open a discussion, and any member can reply. Contributions are reviewed before they appear.
No discussion yet. Open the first thread.
New to MARATTO™? Create a free account.
