Breast Cancer Tumour Classification With Fuzzy Inference System (FIS)
Of all cancers occurring in women, breast cancer is the most common, with 2 million new cases and over 0.5 million deaths globally each year. Many doctors are overwhelmed, having to deal with up to 70 patients a day. Mistakes in diagnosis can have deadly consequences.
This study presents a Python-based fuzzy logic knowledge-based inference system for the binary classification of breast cancer tumours, taking an extensive data-driven, supervised learning approach in defining linguistic terms, implication rule sets and membership functions, with the Wisconsin Diagnostic Breast Cancer dataset used as the empirical, impartial problem domain expert. As a Python-based FIS can be deployed on cost-effective, mobile hardware including raspberry Pi’s it is extremely affordable, portable and readily located in any clinical environment.
A large set of configuration variants that includes 5 defuzzification methods are tested, prioritising the classification accuracy of malignant tumours. The best FIS model classified all test set malignant tumors correctly with an overall accuracy of 93.6%, outperforming the logistical regression, decision tree, random forest and neural network machine learning methods.
Master’s Degree Intelligent Systems paper grade 98% | Source code
Fuzzy Logic Theory
Theoretical assignment.