Deep Learning Algorithm for Predicting the Type of Breast Cancer Using Stack of Multi-Modal Classification
Intisar Mohsin Saadoon, Maha Adham Al-Bayati · 2024
According to the global statistics, breast cancer (BC) which is one of the most prevalent forms of cancer affecting female population across the world ranks amongst the top new cancers cases and cancer death causes in the world today. Therefore, this is one of the most important public health problems for contemporary societies. Accurate classification of benign tumors has significant benefits to the patients in terms of prognosis, likelihood of surviving and prevention against wastage of resources by avoiding an unnecessary treatment. Seven predictive models were applied to analyze a public data set known as CuMiDa_ GSE45827 including K-Nearest Neighbor (K-NN), Support Vector Machines (SVM), Multinomial Naive Bayes (MNB), logistic Regressions (LR), Random Forests (RF), Decision Trees (DT), Gradient Boost (GB), Furthermore, each of these models has been integrated into the stacked Model (SM). In a stack Modal (SM) model, information sources are combined to enhance the classification accuracy performance. SM is a form of multi-modality where it first learns individual models for every modality. Once the models trained they are consolidated into one model that is capable of classifying multi-modal data. Despite being associated with a unique set of challenges; a stack model could offer an exceptionally efficient approach in dealing with seven distinct prediction models. However, the stack model can produce reliable and accurate prognosis for breast cancer if it is addressed by appropriate data preprocessing, model selection, variation, hyperparameter tuning, and evaluation. We used stack model (SM) with the data set and achieved 0.973684 prediction accuracy.