Web application for point of interest recommendation
In the recent years, we have been observed the rising trend of social networks. People have formed a new habit of sharing their experience in every daily activity with their friends online. One of the major type of information that is often found on the social network websites is POI (short for poin...
Saved in:
Main Author: | |
---|---|
Other Authors: | |
Format: | Final Year Project |
Language: | English |
Published: |
2017
|
Subjects: | |
Online Access: | http://hdl.handle.net/10356/70249 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Nanyang Technological University |
Language: | English |
Summary: | In the recent years, we have been observed the rising trend of social networks. People have formed a new habit of sharing their experience in every daily activity with their friends online. One of the major type of information that is often found on the social network websites is POI (short for point of interest) reviews. This is often found in the form of geotagging status/tweet or review in some social networks dedicated for POI review (e.g. Yelp.com, FourSquare.com, …). This kind of information is very valuable, by exploiting it efficiently we can know user’s preference and behavior pattern. We can provide the recommendation service to tell user the POIs that match his/her fondness. On the other hand, people’s preference is extremely useful for the POI owner. Knowing what people like, POI owner can therefore set up the new business that meets people’s need or improve the existing POIs.
There were numerous studies attempted to solving the POI recommendation problem, in this project we will exploit two researches [1, 2] that were conducted in the recent years with the standout approaches. Both two studies used probability models to represent user’s preference and behavior. Study [1] built a model based on sentiment, aspect and region while study [2] built its model based on user’s behavior with respect to time, place and activity. Both two studies had the promising result however they required heavy computation for training data which then limit them from provide recommendation to new user (whose behavior information was not collected in the training data) in sufficient time. In this project, we will attempt to apply this two studies to perform online recommendation. The goal is providing the recommendation in sufficiently short time without re-training the data.
The recommendation service will be provided to user through a web application. There is an existing web application built with Java servlet and Tom Cat server however it is challenging to extend the system due to its monolithic architecture. This project will develop a new system with separation of concern design and micro services architecture. The new system must be robust and maintainable. |
---|