Anticipation of Lung cancer using topical data mining techniques

Abstract

Medical domain provides the most critical and much needed requirements for the machine learning domain. These requirements are crucial due to the fact that the data in medical domain is highly complex in nature, and it requires analysis of multiple and varied types of data to receive at a diagnosis. Cancer identification is one such area, which requires precision and reliability to the maximum extent, due to the high risk involved in the recognition process. Faster and accurate detections become mandatory, as faster detections can prolong the life of patients to a large extent. Cancer is of varied types, depending on the area of occurrence. Hence, the parameter of each cancer type varies significantly with others. Further, as each type of cancer exhibits different requirements, each should be dealt with as independent problems rather than a single recognition system. This thesis specializes on developing models for identifying lung cancer. Lung cancer is one of the most crucial of the cancer disease, as it exhibits the highest mortality levels. Further, the detection process is also complex due to the large number of varied parameters involved in the process. This thesis presents three contributions that effectively handle the challenges in the lung cancer domain and thereby, to provide the best predictions. newlineThe initial contribution provides the baseline for the lung cancer prediction levels. This contribution creates a multi-model structure that can be used for effective prediction of lung cancer. The model uses a combination of Decision Tree and K-Means clustering models. Both the models are integrated at varied levels providing improvements in the prediction process. newline