This paper describes a novel procedure for identifying peak events, starting from the data of presence from mobile phones. The aim is to support the identification of aggregation of peoples with respect to time-series of presences in a given territory. Specifically, the data used in the work consist of a database containing the values of presence of mobile cell phone users with a time interval of 15 minutes. The approach used for the peak identification is based on statistical and machine learning (ML) methods. The study describes the procedure used to recognize a combination of percentiles suitable for the recognition of peak events linked to scenarios related to large events of collective interest. Furthermore, to validate this procedure, supervised learning methods are adopted, namely Logistic Regression (LR) and Support Vector Machines (SVM). Test results confirmed that both methods can recognize peak events with remarkable precision when obtained with the optimal percentile combination method.
Identify peak events from mobile phone presence data
G Felici;C Gentile;G Stecca
2020
Abstract
This paper describes a novel procedure for identifying peak events, starting from the data of presence from mobile phones. The aim is to support the identification of aggregation of peoples with respect to time-series of presences in a given territory. Specifically, the data used in the work consist of a database containing the values of presence of mobile cell phone users with a time interval of 15 minutes. The approach used for the peak identification is based on statistical and machine learning (ML) methods. The study describes the procedure used to recognize a combination of percentiles suitable for the recognition of peak events linked to scenarios related to large events of collective interest. Furthermore, to validate this procedure, supervised learning methods are adopted, namely Logistic Regression (LR) and Support Vector Machines (SVM). Test results confirmed that both methods can recognize peak events with remarkable precision when obtained with the optimal percentile combination method.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.