Fusing Geographic Information Into Latent Factor Model for Pick-Up Region Recommendation
Zhuhua Liao, Jian Zhang, Yizhi Liu, Hao Xiao, Yijiang Zhao, Aiping Yi · 2019
Taxis' trajectories increasingly become an important data source for pick-up regions recommendation. However, long-term and growing taxi trajectory data would inevitably reduce the efficiency of recommendation and consume huge amounts of storage resources. Therefore, a Latent Factor Model integrated with the geographic information, the GeoLFM, is put forward for solving the data sparseness problem caused by the short-term taxis' trajectories. This model makes up the faultiness of data sparseness by integrating the geographic information being related to drivers into the decomposition of matrix. With the comparison between our experimental method and others, the Mean Absolute Error (MAE) between the recommended results and actual values is reduced by 23.4%, while the Root Mean Square Error (RMSE) is lowered by 19.8%.