Читать книгу Bioinformatics and Medical Applications - Группа авторов - Страница 12

Big Data and Machine Learning have been effectively used in medical management leading to cost reduction in treatment, predicting the outbreak of epidemics, avoiding preventable diseases, and, improving the quality of life.

Prediction begins with the machine learning patterns from several existing known datasets and then applying something very similar to an obscure dataset to check the result. In this chapter, we investigate Ensemble Learning which overcomes the limitations of a single algorithm such as bias and variance by using a multitude of algorithms. The focus is not solely increasing the accuracy of weak classification algorithmic programs however additionally implementing the algorithm on a medical dataset wherever it is effectively used for analysis, prediction, and treatment. The consequence of the investigation indicates that ensemble techniques are powerful in improving the forecast accuracy and displaying an acceptable performance in disease prediction. Additionally, we have worked on a procedure to further improve the accuracy post applying ensemble method by focusing on the wrongly classified records and using probabilistic optimization to select pertinent columns by increasing their weight and doing a reclassification which would result in further improved accuracy. The accuracy hence achieved by our proposed method is, by far, quite competitive.

Keywords: Kaggle dataset, machine learning, probabilistic optimization, decision tree, random forest, Naive Bayes, K means, ensemble method, confusion matrix, probability, Euclidean distance