Wavelet Frequency Tensor applied to temporary environments in discrete spaces to obtain mobility patterns. Use case in the detection of routes in a territory. (November 2021)

• Record Type: Journal Article
• Authors: Marchuet, Dani; Palanca, Javi; Botti, Vicent
• Is Part Of: Engineering applications of artificial intelligence. Volume 106(2021)
• Publication Date: 2021-11
• Subjects: Wavelet -- ETL -- Discrete temporal series -- Frequencies tensor -- DBSCAN -- HDBSCAN -- Smart cities; Engineering -- Data processing -- Periodicals; Artificial intelligence -- Periodicals; Expert systems (Computer science) -- Periodicals; Ingénierie -- Informatique -- Périodiques; Intelligence artificielle -- Périodiques; Systèmes experts (Informatique) -- Périodiques; Artificial intelligence; Engineering -- Data processing; Expert systems (Computer science); Periodicals; 620.00285
• DOI: 10.1016/j.engappai.2021.104507
• Languages: English
• ISSNs: 0952-1976

Abstract

Abstract: The use of frequency tensors for the representation of discrete time series information through wavelet transformations offers a methodology that allows the application of classification methods that result in the detection of mobility patterns in a geographical area. The use case focuses on a territory from which the geolocation of anonymized mobile device identifiers is extracted to aggregate its frequencies in order to ensure personal privacy. Furthermore, the kind of datasets we used in our study were treated with other methods to ensure that the identifiers for a same user change over time. Nowadays, this form of data treatment is implemented using a continuous process in smartcity platforms, first by anonymizing and second by changing the seed of the encryption method. Where the suggested technique is applied, a cluster analysis is performed, by which subsets or routes of the movements between the different points studied are obtained. Highlights: We have proposed a technique, that is based on the Wavelet transformation processes to perform a discrete transformation for the extraction of patterns that relate time series of points belonging to any geometric space. The method of obtaining the frequency tensor and the process of generating the extract proposed in the use case achieved strong structures after analyzing the results of the subsequent classification process. We have demonstrated that the proposed methodology facilitates the data analysis process by generating a frequency tensor that represents the discrete Wavelet transformation.