Mobile datasets are widely used as firsthand sources for human mobility research. These datasets are often incomplete or have heterogeneous spatiotemporal resolutions, e.g. a dataset is often aggregated or in lack of fields. In many cases, a reliable dataset in human mobility research comes from sampling or merging original datasets, a challenging task. In this paper, we present our experience on creating a reliable dataset describing mobile data traffic in individual’s spatiotemporal view. We focus on individuals having enough geographical information and merge their call records from one dataset with the data traffic records extracted from another dataset. Based on this dataset, we perform an analysis of user demand on mobile data traffic in terms of spatial and temporal behaviors. For each subscriber, sessions are put into a 3-dimensional space in terms of space, time and volume and are clustered by applying DBScan. Characteristics of are revealed from the statistical analysis on clusters. Subscribers are also categorized according to their clusters.