An Ensemble-Based Approach to Lung Cancer Prediction and Accuracy Improvement

One of the most important and pressing needs in the machine learning field is medicine. Because health-related data is inherently complicated and diagnosis needs analysis of many different kinds of data, these criteria are of the utmost importance. There is a great deal of danger associated with cancer detection, making it an area that demands absolute accuracy and dependability. It is now essential to have quicker and more precise detections since they greatly increase the amount of time patients may live. Many different kinds of cancer may develop in different parts of the body. The detecting method is very complicated because of all the many factors that need to be considered. The three primary ideas presented in this thesis tackle the problems with lung cancer detection in a way that is both effective and efficient. Every component of the bagging method has been fine-tuned to raise the model’s ability to forecast capabilities. In order to provide equal, the bagging method, the bag production method has been adjusted. The data subset that are generated are balanced by using this technique. Effective processing of data with imbalances is aided by this. In order to improve forecasts, the weighted voting approach has been merged with the overall combiner function. Calculating survival rates is quite efficient, according to results and correlations. With a $99 \%$ success rate, the suggested approach outperforms both the decision tree and k-means based survival prediction (DKSP) and Gaussian Classification System (GCS) models. The suggested strategy outperforms DKSP and GCS, which are both respectable, by producing noticeably better results in terms of accuracy.