Date of Award
2-2020
Document Type
Thesis - SCU Access Only
Publisher
Santa Clara : Santa Clara University, 2020.
Degree Name
Master of Science (MS)
Department
Computer Science and Engineering
First Advisor
Yi Fang
Abstract
A pair of latitude and longitude represents only a single point on the earth’s surface, containing zero information about the location. To solve this issue, we developed a universal location encoder capable of generating efficient location representation via numerical vectors, referred to as Efficient and Semantic Location Embedding (ESLE). The ESLE carries both spatial geometry and semantic information, and thus can be used in location information-based applications. To achieve this goal, utilizing the power of deep learning, we first trained a multi-label classifier with Convolutional Neural Network (CNN) models by using the static map tile images, and we extracted ESLE from the CNN models. Then, we developed statistical evaluation methods to evaluate the quality of the obtained location embedding. Finally, to demonstrate the usefulness of the ESLE, we applied them to the task of seeking new ports for NTT DOCOMO’s share-bike business. The results generally verify the effectiveness of the location embedding, and demonstrate a strong potential for its use in location-information based business.
Recommended Citation
Wang, Yuan, "A Deep Learning Approach to Generate Efficient and Semantic Location Embedding" (2020). Computer Science and Engineering Master's Theses. 17.
https://scholarcommons.scu.edu/cseng_mstr/17