Only the correct choice of parameters can make SVM classification better. SVM optimal parameter selection has not been a very good method at present, generally can be solved through cross-test, but this method is quite time-consuming; in practice, grid search method can also be used to determine the parameters, but the method needs to manually determine the parameters Range, change step, and other factors, and the recognition effect is not necessarily ideal. In order to overcome the above defects and avoid the blindness of parameter selection, this paper uses genetic algorithms to achieve parameter optimization in SVM.

Genetic Algorithm GA is a global optimization algorithm based on natural selection and natural genetics. Using GA to operate on a group of multiple entities, the information among individuals can be exchanged through GA, so that individuals in a group can be optimized from generation to generation and gradually approach the optimal solution.

The three main GA operators are: (1) Selection. That is, according to the fitness value of each individual, individuals with higher fitness are more likely to be inherited into the next generation; individuals with lower fitness are less likely to be inherited into the next generation. In this way, the individual's fitness value in the group can be continuously approached to the optimal solution.

(2) Crossover. Also known as reorganization is the selection of two individuals from a group with a large probability and the exchange of one or more of the two individuals. The crossover operator generates children, and the child inherits the basic characteristics of the parent. (3) Variation. That is, one or more bit values ​​on individual code strings are changed with a small probability.

Genetic SVM genetic SVM flow The genetic SVM model proposed in this paper uses GA to optimize parameters in the SVM. The specific process is as follows: (1) system initialization, including parameters and initial population; (2) calculation of the initial objective function value; (3) Determine if the end condition is reached. If so, output the result. Otherwise, go to step (4); (4) Calculate, select, crossover, and mutate the fitness value. (5) Calculate the objective function value of the child and reinsert it. On behalf of the population, go to step (3).

Parameter selection This paper adopts the true value coding method, which is mainly because the true value coding is suitable for genetic search in a large space, and improves the computational complexity of the GA and improves the operation efficiency. Since two parameters C and σ are to be optimized, a random population of 40 individuals, each with 2 variables, is established as the original population. In order to avoid the over-study phenomenon of SVM, the training samples are divided into formal training and auxiliary training samples. The objective function is the sum of the positive rate of the formal training sample set and the auxiliary training sample set.

In the initial stage of genetic search, the positive rate of the formal and auxiliary training sample sets will increase at the same time; as the search progresses, the positive judgment rate of the official training sample set will increase slowly; when the positive training sample set, the positive judgment rate will increase further. When the positive judgment rate of the auxiliary training sample set decreases, the genetic search <7> is stopped.

Conclusion SVM is a theory based on statistical learning, has a good generalization ability, and has been applied more and more in the field of fault diagnosis. This paper uses genetic SVM to perform fault diagnosis of transformers. GA is used to optimize parameters in SVM. This optimization method can quickly and accurately find optimal values ​​in a large range, and it can be further applied to transformer fault diagnosis based on dissolved gases in oil. . The example analysis shows the effectiveness of the proposed algorithm and its superiority over other methods.

Semi-harvester for Agriculture:


The semi-feeding harvester can complete the harvesting, delamination, separation of stems, removal of sundries and other processes at one time, and the Rice Harvester machine for obtaining grain directly from the field is mainly suitable for rice harvesting, wheat harvesting, and this reaper machine can adapt to deep mud feet. Under the serious harvest conditions, the grain cleanliness after harvest is very high, and at the same time, the stem integrity after harvest can be guaranteed, so that farmers can complete harvesting and granulation with a single operation, thus saving manpower and material resources and greatly reducing the burden of farmers.


Semi-harvester for Agriculture02


Semi-harvester for Agriculture Technical Parameters:

1. Size: 3650*1800*1820 (mm)

2. Weight: 1480KG

3. Engine Fuel: Diesel

4. Harvest numbers line: 3

5. Harvesting width: 1200 (mm)

6. Cutting height range: 50-150 (mm)

7. Threshing depth control system: Manually

8. Adaptation crop height: 650-1200 (mm)



If you have any questions, please contact us directly. Crawler tractor for agriculture are produced by Hunan Nongfu with high quality and good appearance.

welcome you can visit our factory for inquiry, please send mail directly to us.




Semi-harvester for Agriculture

China Harvester Machine, Rice Harvester, reaper machine, Agriculture equipment

Hunan NongFu Machinery&Electronic.Co., Ltd. , https://www.nfagmachine.com

Posted on