Code: [ Ссылка ]
University: Stockholm University
Real-world classification problems often present two significant challenges: class imbalance and cost sensitivity. These challenges frequently manifest in scenarios where one class dominates the dataset, such as distinguishing between well-functioning machines and malfunctioning machines, identifying healthy individuals versus those with illnesses, or discerning legitimate work emails from phishing attempts. Recognizing the intrinsic relationship between class imbalance and cost sensitivity in classification tasks, it becomes imperative to develop effective methodologies that address both issues simultaneously. Consequently, our exploration of novel methods for addressing these challenges is motivated by the need to find viable and practical solutions rather than diminishing the value of existing approaches. Our research has yielded a novel approach to simultaneously address the aforementioned problems, exhibiting promising potential for further development and application. Undoubtedly, this assignment has posed considerable challenges and necessitated extensive effort. Nonetheless, the acquired knowledge and experience have significantly strengthened the author's capabilities in the fields of Machine Learning and Deep Learning.
#machinelearning #deeplearning #randomforest #neuralnetworks #imbalance
