Cancer Malignancy Prediction Using Machine Learning: A Cross-Dataset Comparative Study
Priyanshu Rawat, Madhvan Bajaj, Shreshtha Mehta, Vikrant Sharma, Ayushi Jain, Manisha Manjul
Abstract
Cancer is a serious public health problem, and early cancer identification increases the likelihood of effective treatment. In this work, we performed a comparative analysis of four publicly accessible cancer-related datasets to predict cancer malignancy. We compared the performance of four machine learning algorithms: Logistic Regression, Decision Tree Classifier, Random Forest Classifier, and Gaussian Naive Bayes using the Breast Cancer Wisconsin (Diagnostic) Dataset, Lung Cancer Dataset, Prostate Cancer Dataset, and Brain Tumor Dataset. We examined the precision of these models in each dataset, as well as the number of cells, rows, and columns for each dataset. Our findings demonstrate that various machine learning models perform differently on different datasets, and that the optimal model selection is dependent on the dataset being examined. In the Breast Cancer Wisconsin (Diagnostic) dataset, the logistic regression model had the best accuracy, whereas in the Brain Tumor Dataset, the Gaussian Naive Bayes model had the highest accuracy. In the Lung Cancer data set, the Random Forest Classifier model obtained the greatest accuracy. The datasets differed in size and properties, which may have affected the effectiveness of the machine learning models employed for analysis. Our results emphasize the significance of selecting the most suitable machine learning model for a particular dataset and offer insights into the effectiveness of various models in cancer classification.