Recommending third-party APIs via using lightweight graph convolutional neural networks

Figures & data

Figure 1. A motivating example.

Table 1. Statistics of the dataset.

	User	API	Interactions	Density
Original	160309	21031	306751	0.00009
After Processing	6684	21031	121670	0.00087

Figure 2. Comparisons of Recall@K.

Figure 3. Comparisons of NDCG@K.

Table 2. Comparisons of Recall@K.

	5	10	15	20
BPRMF	0.0704	0.0978	0.1280	0.1513
NMF	0.0874	0.1170	0.1460	0.1634
GC-MC	0.0842	0.1120	0.1388	0.1621
NGCF	0.0948	0.1259	0.1561	0.1726
Proposed Method	0.1099	0.1590	0.1933	0.2221

Table 3. Comparisons of NDCG@K.

	5	10	15	20
BPRMF	0.0552	0.0715	0.0821	0.0920
NMF	0.0685	0.0880	0.1018	0.1120
GC-MC	0.0674	0.0915	0.1050	0.1105
NGCF	0.0793	0.1030	0.1182	0.1260
Proposed Method	0.0974	0.1175	0.1292	0.1382

Figure 4. Impact of the embedding size.

Figure 5. Impact of the regularisation coefficient λ.