Spring onion seed demand forecasting using a hybrid Holt-Winters and support vector machine model.

Area of Science:

• Agricultural Economics
• Data Science
• Supply Chain Management

Background:

• Variable spring onion seed demand poses supply chain challenges for seed companies.
• High operational costs stem from long propagation periods and complex logistics.
• Accurate demand forecasting for spring onion seeds has not been previously explored.

Purpose of the Study:

• To develop and evaluate a novel hybrid forecasting model for spring onion seed demand.
• To identify key dynamic factors influencing spring onion seed demand.
• To assess the model's performance against benchmark machine learning approaches.

Main Methods:

• A hybrid Holt-Winters and Support Vector Machine (SVM) model was proposed.
• Dynamic factors including historical sales, inventory, market price, and weather data were used as inputs.
• The hybrid model was compared with two advanced machine learning benchmark models using commercial sales data.

Main Results:

• The hybrid Holt-Winters and SVM model demonstrated superior performance over statistical-based models in forecasting spring onion seed demand.
• Seed inventory, crop market price, and historical sales were identified as significant dynamic factors.
• Seed inventory exhibited short-term influence, while market price and historical sales showed mid-term influence. Absolute minimum temperature was the only long-term influencing factor.

Conclusions:

• The proposed hybrid model offers a promising solution for spring onion seed demand forecasting.
• Understanding the influence of dynamic factors can optimize seed supply chain management.
• The model's methodology can potentially be adapted for forecasting demand for other crop seeds, reducing overall operational costs.