Fuzzy Transfer Learning for Prediction in Data-Shortage and Rapidly-Changing Environments

收集足够的最新数据来训练用于数据分析和预测的学习模型是困难和昂贵的。该项目将开发一种模糊迁移学习方法，使用信息粒度理论，利用其他类似系统中具有不同特征和/或分布的数据，为当前问题提供准确的基于学习的预测。它将建立一个新的研究方向，即预测的模糊转移学习，其成果将使政府和行业能够更好地利用过去的经验来做出更准确的预测和决策。在数据分析、商业智能和决策研究领域也将产生非常显著的效益。

Collecting sufficient up-to-date data to train a learning model for data analysis and prediction is difficult and expensive. This project will develop a Fuzzy Transfer Learning methodology, using Information Granularity theory, that exploits data with different features and/or distributions available in other, similar systems, to provide accurate learning-based prediction for current problems. It will establish a new research direction, Fuzzy Transfer Learning for Prediction, and the outcomes will enable government and industry to better use past experience to make more accurate predictions and decisions. Highly significant benefits will also accrue in the data analytics, business intelligence and decision making research fields.