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

Table 5

Testing AUC comparison of different IdEGs on synthetic datasets.

IdEG types	Dataset: Book-Crossing		Dataset: MovieLens
	Cold-start	Warmed-up	Cold-start	Warmed-up
Rand-IdEG	0.7940 (+0.00%)	0.7943 (+0.00%)	0.7065 (+0.00%)	0.7377 (+0.00%)
Meta-IdEG	0.7945 (+0.06%)	0.7948 (+0.06%)	0.7132 (+0.95%)	0.7680 (+4.11%)
RM-IdEG	0.7951 (+0.14%)	0.7955 (+0.15%)	0.7174 (+1.54%)	0.7735 (+4.85%)

indicates whether the DisNet-NN variant is significantly superior to the coupling algorithm or not on each dataset (pairwise t-test at the 0.05 significance level).