Skip to content

Performed experiments using an ensemble of machine learning (ML) models, including node2vec, for the task of network link prediction in an academic setting.

Notifications You must be signed in to change notification settings

vighneshck/Network-Link-Prediction

Repository files navigation

Network Link Prediction

The submitted code is a single ipython notebook, containing codes for:

(i) loading training graph from train.txt
(ii) visualising the loaded graph
(iii) extracting additional node features from JSON file
(iv) dataset sampling
(v) feature engineering
(vi) displaying feature importance
(vii) Exploratory Data Analysis (EDA) on the engineered dataset
(viii) Reading from test_public.csv
(ix) scaling (optional and model-specific)
(x) training
(xi) testing (and generating kaggle submission)

To open the ipython notebook for execution, type the following on the command line:

$ jupyter notebook Network_Link_Prediction.ipynb

OR, you can alternatively mount the ipython notebook on Google Colab directly.

In the ipython notebook that opens up, run the following sections of the code in the same order specified:

(i) Setting up Google Colab
(ii) Initialise Experiment
(iii) Extracting node information
(iv) Approach 2 (Extra graph features)
(v) Reading from test file for Approach 2
(vi) Standard scaling of all features - OPTIONAL
(vii) Testing for kaggle
(viii) Approach 3 (Node2Vec) - OPTIONAL

About

Performed experiments using an ensemble of machine learning (ML) models, including node2vec, for the task of network link prediction in an academic setting.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published