Research

Hybrid Electric Vehicle Sales Prediction using ANN, ANFIS, and E-Word-of-Mouth

Period: 2023/07/01 to 2024/02/29 (8 months)

National Science and Technology Council (NSTC) Project Number: NSTC 112-2813-C-006-026-H

Summary:

Hybrid Electric Vehicles (HEVs) dominate Taiwan’s electrified vehicle market starting from 2021, yet most sales forecasting research focuses on traditional statistical model and past sales data, leaving a gap in understanding whether appling neural network models and online consumer signals improves the prediction accuracy in HEV demand. The researchers trained both classical Artificial Neural Network (ANN) and an Adaptive Neuro-Fuzzy Inference System (ANFIS) on 16 macroeconomic and raw material indicators , then tested whether adding e-WOM data — sentiment scores from the Mobile01 forum and Google Search Trends — improved forecast accuracy, measured by MAPE. Contrary to prior literature, ANFIS performed substantially worse than ANN (82% vs. 24% MAPE), and incorporating e-WOM did not consistently reduce prediction error. However, this may reflect the limitations of the data source rather than a fundamental weakness of e-WOM itself — Mobile01, while Taiwan’s largest auto forum, captures only a narrow slice of online consumer discourse and may skew toward enthusiasts rather than the broader HEV buying public. Future work incorporating a wider range of platforms — such as PTT, Dcard, or social media — may yet vindicate e-WOM as a meaningful predictor of HEV sales.