基于时空数据的用户社交联系强度研究

Social Strength Learning between Users Based on Spatiotemporal Data

word2vec是Google推出的一款将词表征为实数值的高效开源工具。采用该工具将时空数据中的每位用户表征为一个实数值向量并预测用户间社交联系的强度。提出了在word2vec学习过程中动态调整学习率的算法——Location-weight算法,根据不同位置的不同用户数目在学习过程中加入位置权重,并探索其对用户社交联系强度预测的影响。实验结果表明,加入位置权重的学习算法提高了用户社交联系强度预测的准确性。

Word2vec is a high-efficiency open-source tool issued by Google,which represents a word with a float vector.We used this tool to process spatiotemporal data.Each user is represented with a float vector to predict social strength among users.To predict social strength among users more precisely,this paper proposed a location-weight algorithm that dynamically adjusts the learning rate in the process of learning with word2 vec.According to the number of different users at different locations,we added location weight to the algorithm in the process of learning.Meanwhile,we explored the effect of location-weight on prediction of social strength among users.Experimental results validate performance of the proposed algorithm.