Credit Card Fraud Transaction Detection is one of the most famous financial Data Science Project.
- Importing the required packages into python environment.
- Importing the dataset (Credit Card Fraud Transaction Detection Dataset)
- Preprocessing of data as required.
- Exploratory Data Analysis (EDA) to extract features from the dataset.
- Training of imbalanced dataset using four machine learning classification algorithms and evaluate the trained model using evaluation metrics.
- As the dataset was imbalanced, so I resample the data using SMOTE technique for better accuracy.
- Evaluating the trained model using evaluation metrics.
The dataset used in this problem is taken from Kaggle. The dataset consists of features as V1, V2, V3,....., V28 which are principal components obtained by PCA (Principal Component Analysis), Apart from this the remaning features are "Amount" that contains the total money being transacted and the feature "Class" that contains the labels of fraud or non-fraud trancactions in the form of "1" or "0" respectively.
Model is trained before and after resampling of data using four classification algorithms namely Logistic Regression, Random Forest Classifier, K-Nearest Neighbor(KNN) and Decision Tree Classifier. With these algorithms, the models are trained successfully and evaluate each of the model to find the most suitable model for this problem.
From this, I consider that Random Forest classifier is one of the most suitable model as it gives accuracy of about 99.8% among all the classification algorithms.