Pancreatic Cancer Diagnosis using Swarm Optimization Combined with Machine Learning Techniques

When compared to different types of malignant tumors, pancreatic malignancy is the 12
^th
widespread tumor disease among humans globally. Pancreatic cancer can be identified at the earlier stages using microarray gene analysis. The objective of this work is to classify the gene samples as normal and tumoral through the usage of Particle Swarm Optimization technique along with the supervised machine learning classifiers. A variance-based feature selection is employed to select the most appropriate features and the selected features are transformed using particle swarm optimization to improve the classification performance. Four different machine learning supervised algorithms are realized as classification techniques. When these four vanilla classifiers are considered, the random forest algorithm performs relatively well with balanced accuracy score of 87.5%. This score is improved to 94.4% through the usage of the proposed technique. In addition, the proposed technique enhances the prediction of the other two classifiers as well.