Research On Parameter Optimization Method Of SVM Model For Cell Recognition Based On Python

In the field of medical diagnosis,automatic classification of cell images can effectively improve the accuracy and rate of medical diagnosis.Therefore,the automatic classification of cell images is the focus of current academic research.Due to the complex high-dimensional characteristics of cell images,SVM is an excellent automatic classification algorithm,which can solve the problem of poor high-dimensional characteristics of images.Therefore,it is widely used in the practical application of image classification.The key to realizing cell image recognition classification is to construct a good SVM classifier,and selecting effective SVM kernel function parameters is the main factor for constructing high performance SVM classifier.Therefore,the academic community regards the optimization of the kernel function parameters of SVM as the focus of current research.In recent years,many algorithms have been proposed to automatically optimize SVM kernel parameters,such as grid search algorithms and particle swarm optimization algorithms.The automatic classification of cell images is realized to a large extent,but due to the excessive calculation of these SVM parameter optimization algorithms,the parameter optimization accuracy has not been better,resulting in slower SVM modeling and insufficient recognition accuracy.Therefore,from the perspective of SVM parameter optimization method,this paper proposes a Python-based cell identification SVM model parameter optimization method.The algorithm mainly combines the advantages of Python language to write optimization algorithm and the whole cell recognition classification system function module.The optimization algorithm combines the variable grid search method and the quantum particle group QPSO hybrid algorithm to optimize the parameters of SVM.Firstly,the variable grid search method is used to optimize the parameters.The large grid searches for parameters in a large area and narrows the parameter range.Then the middle grid searches for parameters in the determined area and continues to narrow the parameter range.Then the small grid is used.The determined region continues to search for parameters and narrows the parameter range again.Finally,the quantum particle swarm optimization QPSO algorithm is used to optimize the parameters of the final region,which avoids the accuracy of the optimization parameters of the variable grid search method and the optimization of the quantum particle swarm parameters.The calculation is too large and easy to fall into the local optimum problem,whichgreatly shortens the optimization time,improves the efficiency and accuracy of SVM parameter optimization,and plays an important role in constructing high-precision SVM classifier.The constructed SVM classifier is used to identify the cell image,and compared with the recognition accuracy of the SVM classifier constructed by several traditional parameter optimization algorithms,the validity of the proposed parameter optimization method is verified.