This is a Jupyter Notebook file to run a machine learning demo that implements the Implicit BPR algorithm to recommend courses for Open edX platform.
- Jupyter Notebook from https://jupyter.org/install.html. It will run on Linux (not tested) or MacOS (not on Windows).
- AWS CLI from https://docs.aws.amazon.com/cli/latest/userguide/cli-chap-install.html
Clone the repository to the local computer and run the Jupyter Notebook. At the beginning you will find instructions of how to subscribe to the algorithm in Sagemaker's marketplace. Create a S3 bucket to store all the files. You will also need to create an IAM role with access to Sagemaker and S3.
Update the first cell with actual values of the IAM role and the algorithm ARNs, and the name of the bucket created. Here you can also set the size of the test set.
We have included a dummy file student_courseenrollment.csv that you can use to test,
but will not produce any interesting result.
Replace the student_courseenrollment.csv file in the current directory with actual data.
There is a script in the second cell of the notebook that can be run on an LMS to extract actual data from MySQL.
You can run repeatedly the first cell of the Benchmarking section to check randomly with other users.
This notebook will accumulate the results in the scores.csv and top10counts.csv and will show the results
of these multiple runs in the graphs of the Benchmarking section. Each run will change randomly the testing set.
Make several runs with actual data and see the statistical graphs.
Created by Aulasneo.
This is distributed under Apache GPL license.
For additional questions please contact me at andres@aulasneo.com or find our contact information at our web page.