Deep learning models outperform statistical methods in used electronics price forecasting

A new benchmark study systematically compares deep learning and statistical models for forecasting the resale prices of used electronics. Researchers found that the N-BEATS deep learning architecture significantly outperformed classical statistical methods, achieving a 43% reduction in Mean Absolute Percentage Error (MAPE) at a 365-day horizon. Notably, a single N-BEATS model trained for the longest horizon generalized effectively to shorter forecasting periods, indicating its robustness and stability. AI

IMPACT This research could lead to more accurate pricing for subscription services and resale platforms by improving forecasting models for volatile markets.

RANK_REASON The item is an academic paper presenting a systematic benchmark of forecasting models. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• ARIMA
• arXiv
• deep learning
• LSTM
• N-HiTS
• PatchTST
• price forecasting
• statistical model
• TCN
• used electronics

• paper
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →