Adaptive learning algorithm based price prediction model for auction lots-deep clustering based interval quoting.

Area of Science:

• Machine Learning
• Econometrics
• Data Science

Background:

• Auction item pricing presents challenges due to complex price characteristics.
• Accurate interval price prediction is crucial for market stability and informed bidding.

Purpose of the Study:

• To develop an adaptive learning-based model for precise auction item interval price prediction.
• To address the issue of confusing price classes in auction datasets.

Main Methods:

• A dynamic inter-class distance adaptive learning model was used to identify and cluster confusing price classes.
• A deep clustering algorithm integrating dynamic time warping (DTW) and fuzzy C-means (FCM) was employed.
• A KF-LSTM model combining long short-term memory (LSTM) and dual clustering was constructed.

Main Results:

• The KF-LSTM model achieved an average accuracy of 90.23% and a mean absolute percentage error (MAPE) of 5.41%.
• The model demonstrated an interval coverage rate exceeding 85% for actual auction prices at various confidence levels.
• Significant improvements in prediction accuracy were observed, especially during price fluctuation periods.

Conclusions:

• The proposed KF-LSTM model offers a stable and accurate solution for auction item interval price prediction.
• This research provides a valuable reference for advancing auction price prediction methodologies.