Assignment 9: Tackling the Cold Start Problem

Due Thursday by 3am

Points 100

Submitting a website url

Start Assignment

PREREQUISITES: Review the lectures on Recommender Systems. Pay particular attention to Rule-

based, Collaborative Filtering, and Content-Based Filtering methods, as well as strategies to handle the

cold start problem.

REQUIRED PYTHON PACKAGES (in 'requirements.txt'):

pandas

numpy

scikit-learn

surprise

matplotlib

nltk (optional)

REQUIRED RESOURCES:

storage/u.* (MovieLens [100K] dataset)

The student-resources repository provides scripts to load and explore the dataset. Review the

accompanying storage/README.md.

OBJECTIVES: You will develop three types of recommendation systems: Rule-based, Collaborative

Filtering, and Content-Based Filtering. You will also analyze the cold start problem and propose

solutions. Review the code found in the modules/ directory. The modules included runners (if __name__=

"__main__") as examples to run each module. Review the provided datasets. For this assignment, update

your repository by adding a directory called 'moviemate' in your project root directory.

Task 1: Study and Partition the Dataset

In the module pipeline.py, define a Pipeline class with the following methods:

1. load_dataset(file_path):

Load the dataset using pandas and inspect its structure.

2. partition_data(ratings_df, partition_type=None):

Split the data into training and testing sets using user-stratified sampling and temporal sampling.

Create a notebook notebooks/data_exploration.ipynb to demonstrate:

11/22/24, 9:07 PMAssignment 9: Tackling the Cold Start Problem

https://jhu.instructure.com/courses/82966/assignments/906046?module_item_id=42470521/3

Loading the dataset.

Partitioning the data and visualizing rating distributions and timestamp distributions.

Task 2: Model Comparison

Implement the following in moviemate/modules/:

Rule-based Recommendation (modules/adaptive/filters/rule_based.py)

Recommend top-rated movies overall or by genre.

Collaborative Filtering (modules/adaptive/filters/collaborative_filtering.py)

Implement a User-User and Item-Item Collaborative Filtering model using the Surprise library.

Content-Based Filtering (modules/adaptive/filters/collaborative_filtering.py/content_based.py)

Use genres as item features.

Calculate similarities between movies using TF-IDF and cosine similarity.

Create a notebook notebooks/model_selection.ipynb to demonstrate:

Training each model on the training dataset.

Evaluating each model using metrics such as RMSE/MAE, and nDCG on a test set (See Task 1).

Visualizing performance comparisons and an explanation of the results.

Task 3: Tackle the Cold Start Problem

Create a notebook notebooks/cold_start_analysis.ipynb to simulate and analyze the cold start scenario.

The notebook must:

Simulate scenarios where a new user has rated few and/or no movies.

Explain and offer alternative strategies to mitigate cold starts.

Provide performance analysis and visualizations for each alternative of cold start scenarios.

SUBMISSION: You will need to check in the following files and any supporting python modules:

moviemate/modules/rule_based.py

moviemate/modules/collaborative_filtering.py

moviemate/modules/content_based.py

moviemate/modules/cold_start_analysis.py

moviemate/pipeline.py

moviemate/notebooks/data_exploration.ipynb

moviemate/notebooks/model_selection.ipynb

moviemate/notebooks/cold_start_analysis.ipynb

11/22/24, 9:07 PMAssignment 9: Tackling the Cold Start Problem

https://jhu.instructure.com/courses/82966/assignments/906046?module_item_id=42470522/3

Provide the GitHub URL link to your textwave/notebooks/cold_start_analysis.ipynb file via Canvas

to get credit for this submission.

11/22/24, 9:07 PMAssignment 9: Tackling the Cold Start Problem

https://jhu.instructure.com/courses/82966/assignments/906046?module_item_id=42470523/3