Data science project:
- EDA on the dataset with visually appealing graphs
- Model built:
-
- SVM with grid search for hyperparameters tuning
-
- SVM with feature selection
-
- SVM with SMOTE upsampling on the frauds
-
- Random Forest with grid search for hyperparameter tuning
-
- 3-layer ANN with Keras
- Metrics used: AUC, Precision, Recall
- Sample model results (for 3-layer ANN): AUC 98%, Precision 93%, Recall 74%
- Data: folder containing credit card transactions by Kaggle https://www.kaggle.com/dalpozz/creditcardfraud
- ProgressCheck: preliminary results of proposal, EDAs, inferential stats
- FinalResults: code and graphs wrapped together in jupyterbooks: Part1 for EDA and inferential stats, Part2 for predictive modeling, and a separate report
- [Sklearn]
- [Tensorflow]
- [Keras]
- [statsmodel]
- [matplotlib]
- [seaborn]
- [pandas]
- [numpy]
- [scipy]
- Ginny -- author and organizer of this repo, [contact] (https://github.com/chocolocked)
- Special thanks to Amir Ziai, Jenny Hung for feedbacks on the project