Skip to content

Latest commit

 

History

History
26 lines (19 loc) · 1.49 KB

README.md

File metadata and controls

26 lines (19 loc) · 1.49 KB
Raw

SANSA-POI

This repository contains code to process RDF dataset in SLIPO project which contains a certain amount of Point of Interest(POI). Furthermore, we did some clustering analysis of the data.

Clustering Algorithms

We are mainly using Power Iteration and K-Means clustering algorithms to provide clustering analysis based on the categories of POI. The clustering algorithms are from Spark standard ML library and we developed several encoding methods to encode categorical data to numerical.

  1. OneHot Encoding
  2. Word2Vec Encoding (from Spark ML)
  3. Multidimensional Scaling Encoding (Third Party Library, see pom.xml)

Staypoint Algorithm

Staypoint algorithm determines the stay point based on spatiotemporal data. Stay point is defined as the place where user stays for a while. Stay point is bounded by two parameters i.e., Tmin and Dmax. The minimum time user must be in same place to consider it a stay point is given by Tmin. Whereas Dmax is the maximum distance between two consecutive places. By combining stay point with yelp data, one can determine the interesting venues in a region.

  • Input data:- Spatiotempral RDF data, Tmin and Dmax
  • Output: stay points in a region

Project Management

The project is managed using Maven in order to make it easily runnable in different platform. We provided a run.sh shell script to package and run the project with maven profile dev, and it should be easily adaptable.