Hybrid Predictive Model for Breast Cancer Detection

Abstract

Cancer is a huge concern around the globe. This is a disorder that in many instances is deadly that has impacted many people's lives and will continue to impact many more people's lives. Breast Cancer is the second most cause of deaths in women. While cancer can be avoided and controlled in primary stages, an enormous percentage of patients are very late diagnosed. In one year, 40,000 women die from the disease, a woman died of the disease every 13 minutes. This is much harder to treat early breast cancer diagnosis. This paper presents a hybrid model which is a data mining technique to classify the smallest subset of characteristics that will guarantee a very reliable diagnosis of breast cancer as either benign or malignant in early detection. Naïve Bayes, Support Vector Machine and Random Forest classifiers are performed where they also calculate the time complexity of each of the classifiers. In this paper, the classification of Naïve Bayes is concluded as the best classifier with the lowest time complexity compared to the other two classifiers. Comparison of reliability of these three algorithms by precision, accuracy, recall and f-means, tests high comparison to the other classification algorithm. Such results are very favourable and can be used for diagnosis, prognosis, treatment and recuperation. The overall build hybrid model using ensemble method will be used to predict the cases based on the datasets with much higher accuracy.

 Keywords: breast cancer, classification, complexity, navie Bayes, support vector machine, random forest, highest accuracy