Colon Cancer Detection Using Exhaustive Correlation Feature Selection Based Genetic Decision Support Regression

Many types of cancer are prevalent in humans today, and unfortunately, these cancers are responsible for the death of many people. Cancer is a lethal disease yielded by various genetic and biochemical abnormalities. In the world, Colon Cancer (CC) is the fourth most prevalent cancer to be diagnosed and the third greatest cause of mortality. Because there are no reliable diagnostic indicators and nothing is known about the underlying molecular pathways, the death rate from CC keeps rising. The existing procedure of CC detection is an error-prone and time-consuming process. Also, owing to improper features produced high misclassification. Different stages of cancer modify the expression levels of particular molecules, making early diagnosis and detection challenging. As a result, there is an increased urgency to develop accurate CC detection. To tackle this issue, this paper introduces the Genetic Decision Support Regression (GDSR) algorithm and Exhaustive Correlation Feature Selection (ECFS) for colon cancer detection. The first step involves collecting samples of the colon cancer dataset from the Kaggle repository and normalizing the entire dataset using the Box-plot Normalization process (Bp-Np) technique. The following phase uses the Synthetic Minority Oversampling Technique (SMOTE) approach to determine feature weight. The colon cancer affection rate is then examined using the CC Periodic Influence Rate (CCPIR) approach. Following this, the ECFS algorithm is applied to select suitable colon cancer attributes to reduce the dataset's dimensionality. Finally, the GDSR algorithm is utilized to classify colon cancer based on the best attributes. The proposed algorithm exhibits high classification accuracy, precision, and recall performance. It also reduces the time required for colon cancer detection and has a lower false rate compared to alternative methods.