Description #

The working code behind the transaction fraud detection methodology paper: the shared DataPreProcessing notebook (column selection, categorical encoding, the train test split, and feature scaling) that every classifier notebook imports from, an exploratory data analysis notebook, and all four trained classifiers, Decision Tree, K Nearest Neighbors, Logistic Regression, and Random Forest, each with training, an overfitting check, a classification report, ROC and precision recall curves, a confusion matrix, and a calibration curve. Random Forest is the paper's recommended model for deployment.

Files #

PathSizeSHA-256
files/data-pre-processing.ipynb 15,543 bytes ad77aabc9e117229544ac5645c4e54d0a39eb86d403f5ba1c0eccbcf3028b74a
files/data-visualization.ipynb 3,769,898 bytes 83b40fd55623b5b86148d8c4a371831cf6e5886a581faf8ecbcafcf6e2eb1827
files/decision-tree.ipynb 92,594 bytes 3174eada35120e8dbb6731215b2422624408bd1582e34f63bb29febb3d730a80
files/k-nearest-neighbours.ipynb 97,591 bytes 11779b62174c7f18b50a189d6a493444db472a443046c6f89af024cb1f22012a
files/logistic-regression.ipynb 103,598 bytes 9074d7fd19b670c2181db91aa7cd90bb72eb8b502dc1a1616d85622c676e24f2
files/random-forest.ipynb 92,055 bytes 30a70244a638af1a5f6e340767e49da03a358a8dff5bf0df425fc9a895fae4d7

Each file's SHA-256 is listed above. To confirm a download is unmodified: shasum -a 256 filename

Available to verified institutions at no cost. Verification confirms institutional identity, not eligibility.

Licence: Available to verified institutions

← Back to Initiative 2