Web proceedings papers


Elena Dimitrova and Ana Madevska-Bogdanova


When dealing with cancer patients, doctors often face the dilemma whether to start the patients on a risky and painful treatment. For cases of acute cancers such as Acute Myeloid Leukemia (AML), sometimes it is better to treat only the symptoms, because the cancer treatment might even shorten the patient’s life. Gene expression within cancerous cells can provide more precise overview of the disease and the physical state of the patient. Furthermore, it can be useful for predicting the life expectancy of a patient which will help doctors to make the right decision whether to put the patient on the excruciating cancer treatment. Since gene expression data are very large, their analysis is more efficient and more accurate by utilizing machine learning techniques. This paper proposes new machine learning methodologies for predicting life expectancy of AML patients, integrates gene expression data from medical research for the AML patients and confirms that the data indicating the presence of some gene mutations can be associated with favorable or less favorable outcome of an AML patient. The prepared classification models for the existing datasets showed promising results, with accuracy reaching 92%, and can be used for classification of future AML patients.


AML· Life expectancy · Machine learning · SMOTE · Classification · SVM · Decision trees · Random Forest.