Situation clustering aims at clustering documents based on real world information it contains. This first version is developed for dataset ll_Nepal only. Clustering is based on location and topic information document has.

• All input documents are in json format.
• Topic information comes from field topics of sample data, 9498 out of 29946 documents have this field, totally 11 kinds of topics.
• Location information comes from field geoLocations/geohash, 24834 out of 29946 documents have this field, totally 346 kinds of locations. Sample data also have a field LOC, basically each LOC value is corresponding to a geohash value. But, only 5384 out of 29946 documents have LOC. So, we choose geohash over LOC to represent location information.
• Documents which have exactly same location and have at least one common topic are seen to belong to same cluster.

Git clone package with the following command:

$ git clone https://github.com/dreamysx/situationClustering.git

Execute the following command to run the code:

python [code-file-name] [input-path] [output-path]

For example:

$ python situationClustering.py sampleInput/ sampleOutput/

Then, clustering will be running, and it will tell you how many clusters we got. Clusters can be check in your predefined output directory.