Table 1.

Prior research on human activities and land use based on mobile phone data.

Reference	Analytical Perspective	Feature/Characteristic	Application	Description	Algorithm/Technique
[8,9,11,12,13]	Human activities and mobility patterns	Spatiotemporal call volume: total and hourly call volume managed by each base transceiver station (BTS) (i.e., total number of calls or mobile phone devices managed by a given BTS over a given period)	Classification of urban land use types	These studies have depicted human activity patterns based on extracting spatiotemporal call volume features	FCM [8], NMF [9,13], SVM [11], and k-means [12]
[13,14,15]	Temporal changes in human activities	Temporal call patterns and volume: calculations or estimations of the number of calls or mobile phone devices managed by each BTS tower every hour in a seven-day week (i.e., weekdays and weekends)	Land use detection	These studies have detected land use patterns based on temporal changes in human activities to capture human behaviors’ variation over time (e.g., human activity trough in the middle of the day on weekends)	NMF [13], community detection algorithms [14], and latent Dirichlet allocation [15]
[26]	Human dynamics	Spatiotemporal features: cell tower identification that shows BTSs’ exact location and aggregated mobile network traffic activity for each BTS at 10-min time intervals	Investigation of relationship between human dynamics and land use	This study investigated the correlations between land use and human dynamics, depicting human dynamics as a graph in which nodes are BTS towers and edges represent communication traffic between two nodes	Community detection algorithms
[25]	Human commuting patterns	Users’ daily trajectories based on spatiotemporal features: users’ location represented by cell tower location (e.g., a residence location can be identified based on the most frequently used cell tower locations during the night hours)	Clarification of relationship between commuting flows and variables such as industrial, commercial, residential, and educational land use	This study’s main goal was to gain a fuller understanding of the relationship between land use variables and commuting flows, so a gravity model was used (i.e., a widely used technique for assessing commuting flow patterns), which shows that commuting between two locations i and j, with origin population mi and destination population mj, is proportional to the product of these populations and inversely proportional to a power law of the distance between them [25]	Gravity and regression models
[23]	Human daily and weekly activity patterns	Spatiotemporal call features: spatiotemporal call volume (i.e., total call volume and compared call patterns), communication habits, weekly patterns, and contact features	Land use classification	This study focused on various features to capture many aspects of human activity patterns and depict variation in human activities on weekdays and weekends	RF