Fuzzy-based concept-cognitive learning: An investigation of novel approach to tumor diagnosis analysis

Medical decision-making with high-dimensional complex data has recently become a focus and difficulty in artificial intelligence and the medical field. Tumor diagnosis using data mining technology, from the perspective of gene analysis, can effectively improve the prediction accuracy of patients. For gene databases of tumors with high-dimensional attributes and small sample sizes, tumor classification based on gene analysis is a significant step in the intervention and treatment of tumors. The existing research on the tumor classification of gene data has one prevalent disadvantage: gene obtained via the classification performance evaluation has weak interpretability and universality. This paper presents a concept-cognitive learning system with the three-way analysis (CCL3S) in the fuzzy context for the problem of tumor diagnosis with high-dimensional data, a new fuzzy classifier good at tumor diagnosis. The main steps of the CCL3S include: designing fuzzy recognition to extract the core gene, constructing a fuzzy three-way concept space via the core gene, and finally completing the tumor diagnosis based on the minimum recognition degree. Experimental results on nine tumor gene expression datasets demonstrate that CCL3S achieves better classification performance than some related methods.