Geographic Differential Privacy for Mobile Crowd Coverage Maximization
Authors: Leye Wang, Gehua Qin, Dingqi Yang, Xiao Han, Xiaojuan Ma
AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section, we conduct empirical studies on three real user mobility datasets. We use Algorithm 2 for both single and multi-location coverage scenarios given its practicality (no need to foreknow π). |
| Researcher Affiliation | Academia | 1The Hong Kong University of Science and Technology, 2Shanghai Jiao Tong University 3University of Fribourg, 4Shanghai University of Finance and Economics |
| Pseudocode | Yes | Algorithm 1: Optimal policy for multi-location coverage. Algorithm 2: User selection with dynamic estimating π. |
| Open Source Code | No | The paper does not provide concrete access to source code (e.g., a specific repository link or an explicit code release statement) for the methodology described. |
| Open Datasets | Yes | FS dataset (Yang et al. 2016) contains 1083 Foursquare users check-ins in New York, USA across near one year. D4D dataset (Blondel et al. 2012) includes 5378 users two-week mobile phone call logs with cell tower locations in Abidjan, Cˆote d Ivoire. |
| Dataset Splits | Yes | Among the 45 weeks of user mobility data, we use the last five weeks as the test time period, and first 40 weeks for mobility profiling. (FS dataset) We use the first 18 weekdays for mobility profiling and the remaining four weekdays for testing. (CMCC dataset) We use the first nine weekdays for mobility profiling and the last one weekday for testing. (D4D dataset) |
| Hardware Specification | Yes | It takes about 450 seconds on a commodity laptop with i5-5200U (2.2 GHz), 8G memory. |
| Software Dependencies | Yes | We use Gurobi 7.5 (Gurobi 2014) as the linear programming solver engine to run Algorithm 1 for getting the optimal policy ˆP. |
| Experiment Setup | Yes | Table 1 summarizes the experimental parameters. Notation Values Description ϵ ln(2), ln(4), ln(6), ln(8) differential privacy level δ 0.5, 0.6, 0.7, 0.8 threshold for frequent locations N 1083 (FS), 1315 (CMCC) total number of users 5378 (D4D) α 5% N number of selected users ρ 95% probability for user selection k 6 number of user groups |