딥러닝을 활용한 비타민 수요 예측

The vitamin market has experienced significant growth since the onset of COVID-19. In particular, the demand for high-content vitamins has surged, prompting the launch of various new products. Despite this trend, existing research lacks sufficient studies on demand forecasting using deep learning models. This study aims to predict the demand for high-content vitamins by employing the SARIMA model, an advanced time series forecasting method, alongside deep learning models, such as recurrent neural networks (RNN), long short-term memory networks (LSTM), and gated recurrent units (GRU). Using sales data, this study evaluates the effectiveness of time series forecasting by comparing the performance of the SARIMA model with deep learning-based approaches (RNN, LSTM, GRU). The SARIMA model demonstrates stable demand forecasting by effectively capturing both seasonal and non-seasonal patterns. However, it is limited by its assumption of normal error distribution and its inability to fully account for market volatility. Conversely, deep learning models flexibly handle non-linear characteristics and volatility, with RNN and GRU models particularly excelling in predicting sales fluctuations in the vitamin market. The findings of this study indicate that deep learning models outperform SARIMA model in prediction accuracy, delivering reliable results even in complex market environments. Specifically, LSTM proved effective in capturing long-term dependencies within time series data, while GRU demonstrated an excellent balance between computational efficiency and performance. These results provide valuable insights for strategic decision-making in the vitamin industry, including business planning, new product development, and inventory management. This study also highlights the potential of deep learning models to address the limitations of traditional time series approaches, contributing to enhanced demand forecasting and business data analysis across various industries.