Learning Heterogeneous Spatial-Temporal Representation for Bike-Sharing Demand Prediction
Authors: Youru Li, Zhenfeng Zhu, Deqiang Kong, Meixiang Xu, Yao Zhao1004-1011
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results from Citi-Bike electronic usage records dataset in New York City have illustrated that the proposed model can achieve competitive prediction performance compared with its variants and other baseline models. |
| Researcher Affiliation | Collaboration | Institute of Information Science, Beijing Jiaotong University, Beijing, China Beijing Key Laboratory of Advanced Information Science and Network Technology, Beijing, China Microsoft Multimedia, Beijing, China |
| Pseudocode | Yes | Algorithm 1 STG2Vec (G, H, ω, d, τ, L) |
| Open Source Code | No | The paper does not provide any explicit statements about releasing source code or links to a code repository for the methodology described. |
| Open Datasets | Yes | Citi-Bike Dataset1: We collect the trip data of Citi-Bike system in NYC, from 2017/1/1-2018/5/31 (UTC) as our dataset. ... 1https://www.citibikenyc.com ... Meso West Dataset2: Meso West is an ongoing cooperative project to provide access to current and archive weather observations across the United States. ... 2https://mesowest.utah.edu |
| Dataset Splits | Yes | 9,825 samples selected randomly are used for training and the remaining 2,456 samples are used for testing. Furthermore, when testing the prediction result, we use the previous 12 time intervals (i.e., 12 hours) to predict the bike-sharing demand in the next time interval for each station. ... The first 80% of training samples were selected for training the remaining for parameters tuning. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., CPU/GPU models, memory) used for running the experiments. |
| Software Dependencies | Yes | we transform timeweather attributes into three-dimensional vector by sentence embedding with default setting in Gensim (3.4.0). |
| Experiment Setup | Yes | There are some parameters in STG2Vec, i.e., embedding dimension d, sampling window size ω and epochs τ. Taking into account efficiency and performance, the setting is: d = 3, ω = 10, τ = 50. In addition, we transform timeweather attributes into three-dimensional vector by sentence embedding with default setting in Gensim (3.4.0). Furthermore, we take a single-layered LSTM with size of hidden units: h = 64, batchsize b = 256 and time steps S = 12 which confirmed by grid search and showed in Figure3 partially. |