A Knowledge-Fusion Ranking System with an Attention Network for Making Assignment Recommendations

Table 2

nDCG of ranking methods.

Method	Merit student		Average student		Weak student
	@5	@10	@5	@10	@5	@10
MART [14]	0.701	0.751	0.309	0.319	0.272	0.384
RankNet [15]	0.507	0.547	0.298	0.302	0.433	0.509
RankBoost [16]	0.373	0.318	0.208	0.256	0.369	0.403
AdaRank [17]	0.288	0.362	0.173	0.228	0.196	0.391
CoordAscent [18]	0.466	0.681	0.241	0.293	0.494	0.539
LambdaRank [19]	0.156	0.155	0.235	0.213	0.429	0.495
LambdaMART [20]	0.646	0.749	0.279	0.392	0.515	0.569
ListNET [11]	0.433	0.626	0.191	0.240	0.470	0.464
RandForest [21]	0.731	0.781	0.383	0.366	0.298	0.431
MDPRank [6]	0.344	0.441	0.413	0.541	0.568	0.642
KFRank	0.425	0.444	0.599	0.565	0.636	0.686

Bold values represent the best performance of nDCG@k.