Effective Class-Imbalance Learning Based on SMOTE and Convolutional Neural Networks

Imbalanced Data (ID) is a problem that deters Machine Learning (ML) models from achieving satisfactory results. ID the occurrence of situation where quantity samples belonging to one class outnumbers other by wide margin, making such models’ learning process biased towards majority class. In recent years, address this issue, several solutions have been put forward, which opt for either synthetically generating new data minority or reducing number classes balance data. Hence, in paper, we investigate effectiveness methods based on Deep Neural Networks (DNNs) and Convolutional (CNNs) mixed with variety well-known imbalanced meaning oversampling undersampling. Then, propose CNN-based model combination SMOTE effectively handle To evaluate our methods, used KEEL, breast cancer, Z-Alizadeh Sani datasets. order achieve reliable results, conducted experiments 100 times randomly shuffled distributions. The classification results demonstrate Synthetic Minority Oversampling Technique (SMOTE)-Normalization-CNN outperforms different methodologies 99.08% accuracy 24 Therefore, proposed can be applied binary problems real