This work develops a new density-based clustering scheme, TSS-DBSCAN, which uses DBSCAN and a new method of applying two-phase screening, to reduce the extent of the meaningless expansion of clustering to improve data clustering for numerous related applications. Experimental results demonstrate that the proposed new TSS-DBSCAN scheme has very high noise filtering rate and clustering accuracy (both close to 100%), and is faster than some prominent density-based clustering methods, including KIDBSCAN, DBSCAN, IDBSCAN, QIDBSCAN, and SPY_DBSCAN. The presented approach may be the best density-based clustering method with low time cost in the world currently.