Harnessing Data Mining for Early Detection and Prognosis of Cancer: Techniques and Challenges

Abstract

Cancer is one of the leading causes of mortality worldwide. In 2018, there were approximately 1,735,350 new instances of cancer identified in the United States alone, and 609,640 individuals passed away as a direct result of the disease. Cancers include skin melanoma, lung bronchus cancer, breast cancer, prostate cancer, colon and rectum cancer, bladder cancer, kidney and renal pelvis cancer, and others. Cancer has risen to prominence in the scientific community due to the wide variety of cancers and the enormous number of people it affects. There is still active research on cancer prevention and diagnostic strategies. Using data mining methods, we sought to create a reliable and workable system for cancer diagnosis. Machine learning techniques may assist professionals in creating tools that enable early cancer detection. To improve cancer diagnosis rates, this research aims to introduce a novel machine learning method called the Elephant herding optimized logistic regression (EHOLR) strategy. Histogram equalization (HE) was used for preprocessing the acquired cancer data, and linear discriminant analysis (LDA) was used to extract the data's features. Finally, cancer detection is accomplished using our recommended strategy. The effectiveness of the suggested strategy is then assessed using the performance matrix, namely accuracy, recall, and precision.