Journal of Entrepreneurship & Management

M. Sharmila ¹, R. Punithavathi ²

Affiliations
1 Assistant Professor, Dept. of Information Technology, M. Kumarasamy College of Engineering, Karur, Tamil Nadu, IN
2 Professor and Head, Dept. of Information Technology, M.Kumarasamy College of Engineering, Karur, Tamil Nadu, IN

Abstract

In the current scenario, cancer disease is substantial menace to human life globally. About 32 percent of people worldwide are affected by various types of cancer. But lung cancer depicts the highest ratio. Nowadays peoples are not having awareness about to detect the cancer in early stage. The survival rate of five year for lung cancer disease is 55 percent of the cases are affected most. However, only 14 percent of lung tumor cases are diagnosed at an early stage. For slight tumors the five-year survival rate is simply 3 percent. There are 4 stages in lung cancer. If we predict the disease in I and II stage, it is easy to cure by effortless operations. If it exceeds second stage, it may not be cured. So, diagnosing the cancer in earlier stage is the best solution to predict the patients from death. For that, the system uses the Decision Tree and K-Nearest Neighbor (KNN) Algorithms as preferred classification model. By using these algorithms, it becomes easier to diagnose the cancer in early stage. So, the survival rate of lung cancer patients becomes higher. This comparative analysis, calculates and compares the precision of Random Forest, Naive Bayes and KNN and the preliminary result reveals that ID3 furnish better precision for cancer dataset. The input has been accessed only in numeric format. The algorithms also maintain key stuffs of the dataset, which are predominant for extracting performance, and so it may warrant the correct defense and effective preservation. This leads to protection of any extracting works that depends on the sequence of distances between objects, such as Random Forest, Naive Bayes -search and classification, as well as many visualization techniques. In particular, it establishes a restricted isometric property, in which it is the tight leap on the shrinkage and enlargement of the original distances.

Keywords

Machine Learning; Unsupervised Learning; Naive Bayes classifiers; Decision Tree; Random forest; Decision Support system; Neural network.

Full Text