City-Scale Localization with Telco Big Data

Zhu, Fangzhou; Luo, Chen; Yuan, Mingxuan; Zhu, Yiwen; Zhang, Zhengqing; Gu, Tao; Deng, Ke; Rao, Weixiong; Zeng, Jia

doi:10.1145/2983323.2983345

Cited by 38 publications

(51 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…When given an MR record without position information, machine learning models then predict the associated location. As shown in [37], the authors proposed a contextaware coarse-to-fine regression (CCR) model (implemented by a two-layer RaFs). The CCR model takes as input 258 dimensional coarse features and 34 dimensional fine-grained contextual feature vectors.…”

Section: Background and Related Workmentioning

confidence: 99%

“…Suppose that we are using a RaF regression model to recover the outdoor locations L(s) and L(s ) for the samples s and s , respectively. The outdoor locations are frequently represented by GPS coordinates [12], [19], [23], [37]. Given the two distributed domains D = D , the MR samples s and s within the two domains indicate that the corresponding RNC/CellID and GPS positions are different, indicating s = s and L(s) = L(s ).…”

Section: System Design a General Ideamentioning

confidence: 99%

“…In addition, to generally validate the performance of TLoc in various cities, we collect a 2G GSM MR data set by our app in Urumqi, where only 2G GSM Telco network is available. Since the Urumqi dataset contains a relatively small [13] Classical fingerprinting method No transfer DeepLoc [26] Recent deep neural network method No transfer Non-Transfer [37] Random Forest regression No transfer MTL [27] Multi-task learning in Random Forest No src selection SVR-Transfer [34] Transfer Learning in SVR No src selection TLoc…”

Section: A Datasets and Counterpartsmentioning

confidence: 99%

“…2) The previous work CCR [37] implements a pure RaFbased regression model and has demonstrated better localization accuracy than other existing works including the classic work CellSense. Since CCR does not perform knowledge transfer and TLoc performs knowledge transfer on top of RaFbased regressor, we thus name it Non-Transfer in Table IV.…”

Section: A Datasets and Counterpartsmentioning

confidence: 99%

“…This method suffers from median errors of hundreds and even thousands of meters. More recently, data-driven approaches have attracted intensive research interests in both academia and Telco industry [4], [13], [16], [28], [33], [37]. These approaches leverage geotagged MR samples to build the mapping from MR samples to associated locations, and the mapping is then used to localize the mobile devices in non-geo-tagged samples.…”

mentioning

confidence: 99%

See 4 more Smart Citations

Transfer Learning-Based Outdoor Position Recovery With Cellular Data

Zhang

Ding

Ott

et al. 2021

IEEE Trans. on Mobile Comput.

Self Cite

View full text Add to dashboard Cite

Telecommunication (Telco) outdoor position recovery aims to localize outdoor mobile devices by leveraging measurement report (MR) data. Unfortunately, Telco position recovery requires sufficient amount of MR samples across different areas and suffers from high data collection cost. For an area with scarce MR samples, it is hard to achieve good accuracy. In this paper, by leveraging the recently developed transfer learning techniques, we design a novel Telco position recovery framework, called TLoc, to transfer good models in the carefully selected source domains (those fine-grained small subareas) to a target one which originally suffers from poor localization accuracy. Specifically, TLoc introduces three dedicated components: 1) a new coordinate space to divide an area of interest into smaller domains, 2) a similarity measurement to select best source domains, and 3) an adaptation of an existing transfer learning approach. To the best of our knowledge, TLoc is the first framework that demonstrates the efficacy of applying transfer learning in the Telco outdoor position recovery. To exemplify, on the 2G GSM and 4G LTE MR datasets in Shanghai, TLoc outperforms a nontransfer approach by 27.58% and 26.12% less median errors, and further leads to 47.77% and 49.22% less median errors than a recent fingerprinting approach NBL.

show abstract

Section: Background and Related Workmentioning

confidence: 99%