Skip to content

Latest commit

 

History

History
59 lines (43 loc) · 1.3 KB

README.md

File metadata and controls

59 lines (43 loc) · 1.3 KB
Raw

H2O-Host

Provider of the OrbitDB database for H2O.

Requirements

  • Node 8
  • For periodic data generation (i.e. sample data for testing): Linux, Python 3 + SKLearn
  • For Uber example: Python3, pip3, npm

Run

Install dependencies:

npm install

Hosting sample data for testing (Linux only)

./periodic_data_gen

Hosting the Uber example data

Using our example:

cd uber_example
node host

If you get a 'No OrbitDB database found' message in H2O, comment out the

Generating a new sample of pickup data and hosting it:

cd uber_example
pip3 install wheel pandas matplotlib subsample
npm install orbit-db ipfs
subsample -n [YOUR_NUMBER_OF_SAMPLES] uber.csv > uber_sample.csv
python3 uber.py
node host

200 generally works well for this dataset but feel free to experiment - there be dragons.

Using your own dataset

Put a JSON with the following format in the root directory:

{
  "x": [0.2,8.5,...]
  "y": [-3.4,12.2,...]
  "t": [0, 1, ...]
}

The t field is ground truth, i.e. the clusters. This is optional, simply remove the line await db.put( { _id: 't', array: data.t }) in host.js if you don't have ground truth.

Advanced: if you're familiar with SKLearn, any dataset that can be clustered with sklearn.kmeans will work as long as it is in JSON format.