A machine learning project for classifying bank marketing campaign success. This project includes data preprocessing, model training, and evaluation using techniques like SMOTE, ADASYN, and more to handle class imbalance.
- Data Preprocessing: Cleaning and transforming raw data into a suitable format for model training.
- Handling Imbalanced Data: Utilizing techniques such as SMOTE, ADASYN, and undersampling to balance the dataset.
- Model Training: Implementing various machine learning algorithms including MLPClassifier, Random Forest, Decision Tree, KNN, and SVM.
- Evaluation: Assessing model performance using metrics like accuracy, precision, recall, and F1 score.
The dataset is sourced from the UCI Machine Learning Repository. It includes details on customers' bank accounts and marketing interactions.
The dataset is highly imbalanced with approximately 88% 'no' responses and 12% 'yes' responses. This imbalance necessitates the use of techniques like SMOTE and ADASYN to balance the classes. After balancing and training various models, the best results achieved are:
-
Best Accuracy: 0.90
-
Best F1 Score: 0.50