Human mobility prediction with causal and spatial-constrained multi-task network

Huang, Zongyuan; Xu, Shengyuan; Wang, Menghan; Wu, Hansi; Xu, Yanyan; Jin, Yaohui

doi:10.1140/epjds/s13688-024-00460-7

EPJ Data Science

Table 2 Performance comparison between baselines, CSLSL, its variants and ablations on three real-world datasets

From: Human mobility prediction with causal and spatial-constrained multi-task network

	NYC						TKY						Dallas
	Category			Location			Category			Location			Location
	R@1	R@5	R@10	R@1	R@5	R@10	R@1	R@5	R@10	R@1	R@5	R@10	R@1	R@5	R@10
FPMC-D	0.217	0.510	0.631	0.173	0.424	0.526	0.462	0.623	0.688	0.176	0.408	0.503	0.086	0.204	0.261
FPMC-W	0.196	0.394	0.482	0.162	0.323	0.382	0.428	0.513	0.557	0.117	0.245	0.313	0.055	0.125	0.161
DeepMove	0.244	0.499	0.586	0.198	0.407	0.470	0.398	0.567	0.626	0.161	0.331	0.400	0.089	0.185	0.229
Flashback^*	0.280	0.618	0.737	0.223	0.521	0.639	0.479	0.730	0.801	0.207	0.486	0.583	0.090	0.199	0.256
LSTPM^*	0.335	0.659	0.765	0.267	0.560	0.662	0.461	0.727	0.797	0.231	0.457	0.543	0.123	0.247	0.316
GeoSAN	0.193	0.433	0.602	0.166	0.430	0.584	0.317	0.551	0.694	0.158	0.392	0.528	0.078	0.186	0.265
STAN	0.218	0.480	0.591	0.192	0.411	0.494	0.376	0.575	0.668	0.167	0.388	0.478	0.074	0.153	0.196
GETNext	0.303	0.646	0.749	0.246	0.536	0.622	0.452	0.758	0.844	0.216	0.456	0.550	–	–	–
LSL	0.288	0.587	0.684	0.242	0.506	0.589	0.446	0.697	0.762	0.225	0.462	0.548	0.101	0.203	0.254
SBLSL	0.290	0.595	0.682	0.242	0.488	0.571	0.446	0.769	0.851	0.229	0.460	0.545	0.107	0.194	0.234
MELSL	0.275	0.586	0.662	0.227	0.432	0.526	0.439	0.648	0.745	0.211	0.419	0.475	0.078	0.121	0.164
SLSL	0.281	0.611	0.722	0.253	0.534	0.632	0.409	0.731	0.825	0.233	0.458	0.559	0.114	0.220	0.272
HLSL	0.296	0.628	0.734	0.256	0.536	0.625	0.441	0.760	0.847	0.232	0.472	0.562	0.115	0.230	0.284
CLSL-ctl	0.315	0.647	0.745	0.257	0.543	0.636	0.472	0.780	0.861	0.227	0.472	0.560	–	–	–
CLSL	0.322	0.658	0.747	0.261	0.553	0.643	0.459	0.782	0.864	0.238	0.476	0.567	0.120	0.229	0.282
CSLSL-c	0.315	0.638	0.757	0.247	0.546	0.643	0.450	0.775	0.858	0.230	0.463	0.551	–	–	–
CSLSL-t	0.319	0.648	0.749	0.264	0.556	0.652	0.479	0.740	0.829	0.233	0.478	0.568	0.118	0.231	0.284
CSLSL	0.327	0.661	0.759	0.268	0.568	0.656	0.488	0.801	0.875	0.240	0.488	0.580	0.126	0.243	0.297

^*Flashback and LSTPM filter out much more sparse users in their data preparation, reducing the challenge of prediction.

Back to article page