Ανάλυση κλινικών δεδομένων με χρήση των αλγορίθμων μηχανικής μάθησης για την πρόβλεψη του καρκίνου
View/ Open
Keywords
Εξόρυξη δεδομένων ; Σύγκριση αλγορίθμων μηχανικής μάθησης ; WBC dataset ; Αλγόριθμοι ταξινόμησης ; Αναγνώριση προτύπων ; Τεχνητά νευρωνικά δίκτυα ; Μπεϋζιανά δίκτυαAbstract
Cancer has proven to be the plague of our time, as it is being placed by the statistics as the second cause of death worldwide. Although early cancer detection and diagnosis are major steps for fighting the disease, cancer prognosis remains equally important. Prognosis relates with the prediction of the likelihood of cancer, as well as the prediction of a patient’s relapse or survival (life expectancy). All the above have been the object in several recently published studies. The researchers have attempted to develop integrated clinical decision support systems by using machine learning methods and classification algorithms. These systems are able to produce an accurate prediction of the patient’s outcome, which may help clinicians in personalized decision-making.
The application of machine learning algorithms is now the new reality in modern data analysis methods. The large number of methods and the parameters that can be applied in each case allows them to be flexible and at the same time make the analyst more careful in his choice. This choice depends on many parameters such as the content of the dataset and is the key factor for extracting precision results.
A comparative study was performed to evaluate the most characteristic and modern algorithms used in data mining. The comparison concerned the detailed examination of previous investigations with the help of meta-analysis and their application to clinical data with the help of statistical software R. The results of the study agreed with those of previous investigations and confirmed the use of algorithms on a case-by-case basis, but also depending on how the methods are measured accurately. Among the highest-precision algorithms were NN and ΚNN.