A fuzzy data mining approach for remote sensing image recommendation

Nowadays research on Remote Sensing Images (RS-Images) ranking and recommendation for meeting the user-specific Area-Of-Interest (AOI) has received a log of attentions due to a wide range of potential applications. In this paper, we propose a novel approach named Fuzzy rs-Image Recommender (FIR) to rank and recommend relevant RS-Images according to the queried AOI. In FIR, we first propose two features named Available Space (AS) and Image Extension (IE) as two indicators to represent the relationships between AOI and RS-Image. Then, we mine the fuzzy association rules between the proposed indicators and user rating score. Finally, we propose two fuzzy inference strategies named FIR with Weight_area (FIRarea) and FIR with Weight_all(FIR all) to rank and recommend the relevant RS-Images to users. To our best knowledge, this is the first work on RS-Image recommendation that considers the issues of feature extraction and fuzzy rule mining, simultaneously. Through comprehensive experimental evaluations, the results show that the proposed FIR approach outperforms the state-of-the-art approach Hausdorff in terms of Normalized Discounted Cumulative Gain (NDCG).