A recommender system for kiva.org micro-loans based on collaborative filtering.
- TODO
The project consists of three jupyter notebooks:
- kiva-cf
- Initial data cleanup and formatting.
- Introductory exploration of implicit framework capabilities and application to the KIVA dataset.
- kiva-cf-polara
- Translation of done work to polara framework.
- Data reformatting.
- Creation of an evaluation environment.
- result-analysis
- Analysis and visualization of evaluation results used for interpretation.
To start the notebook locate inside the project root directory and run:
jupyter notebook
Full data snapshot can be found here.
All needed csv files should be located in /additional-kiva-snapshot directory.
All pickle files should be located in /pickle directory.
Evaluation results are located in /eval_results directory.