Evaluating the role of different artificial intelligence algorithms in increasing the speed of neural networks used in pattern recognitionof breast masses

Background and aims: There are different types of breast masses that are formed in the breasttissue. In some cases, a breast masses indicates breast cancer. Based on tissue characteristics, sizeand volume, they are classified into benign and malignant types.Artificial Neural Network (ANN) is a complex system based on the function of the human brainand its nervous system. With the increasing use of artificial intelligence in medical sciences,ANNs can be used as a useful method in pattern recognition for the automatic diagnosis of cancerand determining the type of glands.The purpose of this paper is to optimize the performance of the perceptron neural network inincreasing the accuracy of diagnosing malignant or benign cancerous tumors. Different artificialintelligence algorithms are examined in the optimization of the initial weights of the MLP network,and the best optimization algorithm that has a high CCR is selected.Method: The dataset used in this article is related to the cancerous mass information of ۶۹۹people with breast cancer, for each cancerous mass, ۹ characteristics related to size, texture, compression,etc. have been obtained. Based on these characteristics, patients are divided into benignand malignant groups.The research method used in this article is a practical method that can be used in the field of medicineand the development of medical equipment by using artificial intelligence in the automaticdiagnosis of the type of glands.To check the optimization in determining the initial weights of the perceptron network, Particleswarm optimization algorithms and the bee algorithm have been used. Swiping the parametersthat can be adjusted, the results are collected and checked. Feature selection is done by BinaryPSO algorithm.Results: At first, the MLP neural network was formed without optimization and with an accuracyof ۹۴.۲۶%. BPSO feature selection method was used to identify the effective feature. The classificationaccuracy increased to ۹۶.۱۶%. In the optimization of the network with the PSO method,the best answer of accuracy equal to ۹۸.۵۷ was obtained. by using BEE method obtained ۱۰۰%accuracy.Conclusion: The PSO method is better than the BPSO method in terms of detection power, on theother hand, the time rate in the BPSO method is more reasonable because the completion time ofthe program increases with the increase in the population. But the BEE method is superior to thePSO and BPSO methods in terms of accuracy, and it is more logical in terms of time. By comparingthe obtained results, we conclude that the repetition of high correct detection percentagein BEE algorithm is more than other algorithms. The perceptron neural network optimized withartificial intelligence algorithms was able to classify cancerous tumors into benign and malignantgroups with ۱۰۰% accuracy.

کلیدواژه ها:

Breast cancer ، Perceptron neural network ، Particle swarm optimization algorithm ، Bee algorithm