**7. Conclusions**

In this paper, we propose a framework consisting of spatial and temporal features to predict long-term rental/drop-off demand in newly established stations, e.g., in expansion areas. Specifically, we extract features from multi-source open data, propose G-clustering, and apply regression models to predict the demand of stations in three batches according to the established periods. Experiments carried out in the New York Citi bike sharing system demonstrate that our framework for long-term prediction in expansion areas is applicable and outperforms baselines. In the future, we aim to analyze more factors, such as transfer probability from downtown to the suburbs and deal with unusual events to improve predicting accuracy.

**Author Contributions:** Supervision, H.-P.H.; methodology, H.-P.H., F.L. and T.-Y.K.; validation, T.-Y.K.; investigation, H.-P.H., F.L., J.J. and T.-Y.K.; writing—original draft preparation, F.L., J.J. and T.-Y.K.; writing—review and editing, H.-P.H. and Y.-E.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Ministry of Science and Technology (MOST) of Taiwan under grants MOST 108-2221-E-006-142, MOST 108-2636-E-006-013, and MOST 109-2636-E-006-025 (MOST Young Scholar Fellowship).

**Acknowledgments:** This work was partially supported by the Ministry of Science and Technology (MOST) of Taiwan under grants MOST 108-2221-E-006-142, MOST 108-2636-E-006-013, and MOST 109-2636-E-006-025(MOST Young Scholar Fellowship).

**Conflicts of Interest:** The authors declare no actual or potential conflict of interest.
