The exponential growth of big data presents a significant task for the incremental clustering problem. This paper proposes a density-based total clustering method in network space (NS-IDBSCAN) to cluster newly generated data. Instead of re-clustering the entire massive database, which would be much larger than the new datasets, we perform clustering with only the added data. Based on the current clustering results of the old data, the proposed algorithm checks the role of each newly added point and its neighbors in performing clustering. Depending on the density, an added point can be classified as noise, border, or core. This approach dramatically reduces the response time. Moreover, using a hash table with the key as the element's index corresponding to the point Id to store the cluster Id eliminates the need for searching by direct access, further increasing processing speed. Additionally, the proposed algorithm improves accuracy by reducing the intra-cluster distance when eliminating the phenomenon that the border points may belong to more than one cluster, as presented in the discussion. The proposed algorithm is compared with the spatial data clustering algorithm in network space. Test results on three data sources downloaded from OpenStreetMap, ESRI Open Data, and Inside Airbnb demonstrate that the proposed method significantly speeds up the processing time. The accuracy of the proposed algorithm is measured using eight indicators: Silhouette, BSS, WSS, WB, Davis-Bouldin, Dunn, Calinski-Harabasz, and NMI.