论文笔记

Using GPS to learn significant locations and predict movement across multiple users

一些表达

1. In order for any predictions we make to be meaningful, we want to discard as much of the data as possible. It would be quite useless to tell the user, ‘‘You’re currently at 33.93885 N, 84.33697 W and there’s a probability of 74% that you’ll move to 33.93885 N, 84.33713 W next.’’ Instead, we would like to find points that have some significance to the user and perform predictions with those.
2. Instead, reminding the user several miles in advance, or even as she enters her car, would be more productive.
3. Reminders are not the only possible use for the single user: wearable computer systems issues may be addressed as well.

思路

利用马尔科夫模型，把数据聚类（文中使用K-means）得到位置，把每个位置当做马尔科夫中的node，计算到达每一个节点的概率。
疑问：？？
？ 马尔科夫模型具体怎么实现
？ 计算概率怎么预测下一个位置，是否意味着预测结果只能从访问过的位置中产生。

Predicting the Next Location: A Recurrent Model with Spatial and Temporal Contexts

一些表达

1. With the rapid growth of available information on the internet and the enhancing ability of systems in collecting information, more and more temporal and spatial contexts have been collected.
2. Factorizing Personalized Markov Chain (FPMC) (Rendle, Freudenthaler, and SchmidtThieme 2010) is a personalized extension of common markov chain models, and has become one of the most popular methods for sequential prediction. FPMC has also been applied for next location prediction(Chengetal.2013; Chen, Liu, and Yu 2014).
3. In this paper, to better model spatial and temporal information, we propose a novel method called Spatial Temporal Recurrent Neural Networks (ST-RNN）