Make a restaurant recommendation engine for Zomato using user review data.
As a part of the course requirement, we will be conducting Assignment 2 for 20 Marks. This assignment requires you to build a Recommendation Engine for Zomato Media Pvt Ltd and has to be done independently.
This assignment is divided into two parts(10+10 Marks). Part Two can only be done if you complete Part One.
Part One: Implement a basic autoencoder recommender system.
- This part is like the last assignment, here we want you to complete the given jupyter notebook and submit it to us before the deadline.
- The notebook contains some function definitions, you need to complete. Further details are mentioned in the notebook itself.
- The framework used here is PyTorch, but most functions are simple and can be looked up in the Workshop Tutorials and on the internet. For your convenience, we have also implemented a simple threeLayerNet in the notebook for better understanding.
- The notebook and the related data files can be downloaded from this link. Please ensure you download all the files correctly.
- Needless to say, the assignment will be evaluated and marks have been mentioned for each task. All the submissions will be checked for plagiarism.
Part Two: Kaggle Competition
- Once you have implemented the basic model, you are required to improve it and obtain a better score(loss) for the test data.
- This is an open-ended problem and you need to submit your predictions on Kaggle as a CSV file. A function for generating the file from predictions is given at the very end of the notebook from Part One.
- Here we do not restrict you to use your framework of choice. You are allowed to port the given code to whichever framework you feel comfortable with.
- We have also provided with a list of suggestions you can try out to improve your model and obtain a better score.
- Finally, you will be marked on your final score on Kaggle, for which you also have to submit a source file which can reproduce the same score. Marking here will be relative, given the source file is okay, the top 10 percentile will get 10 Marks and the next 10 percentile 9 marks and so on. Though the marking scheme is not fixed as of now, we will update you with the marking scheme later.
- Link to the Kaggle competition is: https://www.kaggle.com/t/53e638db417b4b15aaadc9f0551c3118
Deadline of this assignment is 7th April 2019 11:59 pm.