Deep Interest-Shifting Network with Meta-Embeddings for Fresh Item Recommendation

Table 6

Testing AUC comparison of different IdEGs on the Taobao-Fresh dataset. The best ones are shown in bold.

IdEG types	Cold-start	Warmed-up
Rand-IdEG	0.5792 (+0.00%)	0.6042 (+0.00%)
Meta-IdEG	0.6133 (+5.89%)	0.6361 (+5.28%)
RM-IdEG	0.6160 (+6.35%)	0.6382 (+5.63%)

indicates whether RM-IdEG is significantly superior to the coupling algorithm or not (pairwise t-test at the 0.05 significance level).