rishusiva / turing_twins_hackrx_3.0

Bajaj HackRx 3.0 Finalists

Pre-processing Amazon SNAP user data and meta data. Predict co-purchases by the customer as to enhance buying and selling of goods on retail websites like Amazon. Also recommend items for the purchases made.

We are trying to solve the problem associated with understanding the huge Amazon SNAP dataset and thereby pre-processing it to recommend the co-purchased items.

We have a two way solution:

Recommendation systems for individual items. The meta data file contains 4 features for which we are able to predict the co-purchased items using graph based and analytics. In doing so we have also analyzed that, people buying books only buys books and same with other items. Thus our recommendation system suggests items of the same category and needed a recommendation system for the whole dataset, thus our second solution.

Recommendations based on community predictions using directed subgraphs of SNAP dataset with Transaction data and Meta Data, using Girvan Newman Algorithm. Now we have directed graphs, using which top and next level communities are predicted.

So when Amazon uploads transaction data of any time frame, using our approach they can predict the frequently co-purchased items and focus more on highlighting those items to the target customer.

All the pre-requisites can be installed using pip install -r requirements.txt

Python

 - nltk
 - regex
 - networkX
 - packages like numpy, pandas etc

• Clone the repository using :

    $ git clone https://github.com/HackRx3/PS2_Turing_Twins
  

• Enter the directory using:

    $ cd PS2_Turing_Twins/
  

• Install the requirements using:

    $ pip install -r requirements.txt
  

• To use our Recommender Systems, run the correct ipynb file

• Turing Twins