The advances in data collection due to automated patient monitoring, and the use of electronic patient records, are making it possible to develop computational models that will help the specialists in making their prognoses. In this project, the records of patients who had been treated in an intensive care unit were used in testing different machine learning approaches.
Machine learning is the area of artificial intelligence that involves the development of these techniques for enabling computers to acquire knowledge. The methods work by determining features of a record, in this case a patient record, and training a machine learning algorithm to distinguish the classes using the features. The trained method can then be applied to similar datasets. The success, or failure, of the method depends on the choice of features and on the machine learning algorithm.
Many features were chosen for use in this project, including weight, age, white blood cell count, blood glucose, temperature and haemoglobin level.
Any of a large number of machine learning methods could be applied to the classification problem. In this project the support vector machine, artificial neural network, decision trees and random forest approaches were tested.
|Distance learning in computational biology||Research projects in computational biology|