Cervical cancer prediction using outlier deduction and over sampling methods

Cervical cancer is one of the disease considered to be fourth among the most common types of cancer in women around the world.

The early deduction of cervical cancer helps to raise number of recovery patients and reduce death rates. This research work aims to

use machine learning algorithms to predicting cervical cancer with high accuracy using an outlier deduction and over sampling

method. Analyse the cervical cancer data available the dataset from UCI repository. In this research, first step removes outliers by

using outlier detection method such as density-based spatial clustering of applications with noise (DBSCAN) and by increasing the

number of cases in the dataset in a balanced way through the synthetic minority over-sampling technique (SMOTE) and SMOTE

with Tomek link (SMOTETOmek). Finally, it employs random forest (RF) as a classifier to check the accuracy. Thus, the prediction

model Have a two scenarios: (1) DBSCAN + SMOTETomek + RF, (2) DBSCAN + SMOTE+ RF. I found that combination of

DBSCAN with SMOTE provided better performance than DBSCAN with SMOTETomek. Also observed that RF performed the best

among several popular machine learning classifiers. Furthermore, the proposed research work showed better accuracy than

previously proposed methods for forecasting cervical cancer.