Application Research Of Intelligent Algorithm In The Prediction Of Water Quality In Aquaculture

In recent years, with the country’s strong support for the aquaculture industry, intensive industrialaquaculture has been developing quickly, and has gradually become one of the main aquaculture.Aquaculture water quality which is good or bad plays a decisive impact on fish growth, and with theincreasing density of breeding for aquaculture water quality prediction techniques are increasingly highrequirements. Research aquaculture water ammonia content forecast accuracy problem, you can provide ascientific basis for guidance aquaculture. Due to the highly coupling factors of aquaculture water quality,but it is difficult to establish a universal model to accurately predict. In this paper, commonly used topredict water quality prediction method first introduced, and its comparative advantages and disadvantagesof doing the analysis and comparison, the final choice has a computing speed, strong marketingcapabilities least squares support vector machine for aquaculture water quality prediction.Support vector machine (SVM) based on statistical learning theory is a new machine learningalgorithm. It follows the principle of structural risk minimization, can solve the small sample, nonlinear inthe condition of common falls into local optimization and learning problems such as machine learningtradition, has strong generalization ability. LS-SVM is an extension of SVM based on the constraintcondition by the inequality will be changed into the equation, which avoids solving two quadraticprogramming problems, able to solve model analysis to get. The performance of LS-SVM depends largelyon the choice of its parameters, and inappropriate parameters are often difficult to achieve the ideal effect inthe prediction of. However, at present the method of selecting parameters is more dependent on experience,low efficiency.Quantum behaved particle swarm optimization (QPSO) algorithm as the improved particle swarmoptimization (PSO) algorithm, has faster convergence speed, robustness and better, more and moreparameters to be applied to the optimization of LS-SVM. But as a new optimization algorithm, manyaspects are not perfect. So in order to local search and global search ability of QPSO algorithm in order toachieve a better balance, improve the comprehensive performance optimization, proposed that quantumparticle self-adjustment strategy of a composite weight swarm optimization (ACWQSPO) algorithm, usingthe composite strategy contraction expansion coefficient, the convergence precision and robustness areobtained a certain degree of improvement. While the sample data set for modeling inevitably exist errorsand the resulting effect on the model performance, proposes a weighted least square Laplace distributionfunction of support vector machine (LWLS-SVM). The new algorithm makes use of the characteristics ofthe Laplace distribution, and reduces the side effects of singular sample on the model performance, makingits sparse and robustness to be strengthened.Finally, the article lists the main factors affecting water quality in aquaculture, and analyzes therelationship between them, and selects a significant impact on water quality as a predictor of the ammoniacontent objects. For sample data collection process will inevitably lead to errors in question, the use ofthese data in the first cluster analysis of them, excluding the singular data. This paper studies thecharacteristics and parameters affecting the participation of WLS-SVM prediction of water quality data forthe sample regression prediction accuracy, the choice ACWQPSO parameter optimization WLS-SVM is proposed ACWQPSO optimization WLS-SVM prediction model, and the model is used to predictaquaculture water quality prediction in a certain area. Through the experimental comparison analysis,demonstrates the feasibility of application of prediction model set up in the paper for the aquaculture waterquality prediction, has good practical value.