Research On Artificial Immune Algorithms And Their Applications In Cancer Diagnosis

Artificial Immune System(AIS) is type of adaptive system inspired by principles and mechanisms of the natural immune system. For the last decade, AIS has become a research focus and been applied in various fields. In this paper, we combine data mining with AIS to explore the application of AIS algorithm in cancer diagnosis and classification. The main work of this paper are summarized as follows:Firstly, we summarized the biological foundation of AIS, introduced common classification algorithms and their evaluating index.Secondly, we discussed Clonal Selection Algorithm (CSA) and Clonal Selection Classification Algorithm (CSCA), and applied them in study for cancer diagnosis. By comparing the experimental results and the classification algorithm performance, we concluded that CSCA has higher accuracy than other traditional classification algorithms. Then we explored the changing situations of classification accuracy with different parameters.Thirdly, algorithm principles and processes of the Artificial Immune Recognition System (AIRS) and its subsequent versions (AIRS2, Parallel AIRS) were expatiated in detail. Based on the cancer cases dataset, we made classification experiments and obtained some data. Comparing the classification performance, we found that AIRS1 algorithm is advanced than the other two classifiers. Then we made comparison between these three classifiers and the traditional ones, and explored the influences of different parameter settings on the accuracy.Finally, according to Immune Negative Selection Theory, we presented a Negative Selection Classification Algorithm on the basis of Clonal Iteration (CI-ASCA). Then we thoroughly introduced this algorithm's principles and process and analyzed the classification performance data. The algorithm was proved to be feasible by comparing with traditional classification algorithms and immune algorithms mentioned before. Afterwards, we investigated the change trend of classification accuracy with different parameters and contrasted two classification algorithms which are based on distance judgment and iterative judgment separately in accuracy. We also made comparison between the algorithms of AIS above.