基于PSO-Stacking的河蟹投饵量预测模型

Prediction Model for Feeding Amount of River Crab Based on PSO-Stacking

摘要:

河蟹作为我国重要的水产养殖物种之一，深受消费者喜爱，在河蟹养殖过程中，科学的投饵量是保证河蟹健康生长及提高养殖效益的关键因素。本文通过综合分析影响河蟹养殖投饵量的多种因素，采用集成学习算法建立河蟹养殖投饵量预测模型。搭建数据采集系统，采集包括河蟹生物量、河蟹数量、性别比例、水体pH值、温度、溶解氧含量以及河蟹摄食量等关键参数数据，建立投饵量数据集;运用数据预处理技术对数据集进行平滑处理以及归一化，减少异常值对预测结果的干扰，同时消除特征数据不同量纲的影响;引入粒子群优化算法改进集成学习，建立了河蟹养殖投饵量预测模型，实现河蟹养殖投饵量的准确预测。实际应用测试结果表明本文模型平均绝对误差为0.34971 g，均方根误差为0.49114 g，决定系数达0.903 58。

Abstract:

As one of the important aquaculture species in China, river crabs are well-loved by consumers.In the process of river crab aquaculture, scientific baiting is a key factor to ensure the healthy growth of river crabs and improve aquaculture efficiency. By comprehensively analyzing the factors affecting the baiting amount of river crab aquaculture, an ensemble learning algorithm was used to establish a prediction model for the baiting amount of river crab aquaculture. A data collection system was set up to collect key parameters such as river crab biomass, crab population, sex ratio, water pH value, temperature, dissolved oxygen, and crab feeding amounts to establish a baiting data set;data preprocessing techniques were used to smooth and normalize the data set to reduce the interference of outliers on the prediction results, and at the same time to eliminate the influence of different scales of the characteristic data;the particle swarm optimization （PSO） algorithm was introduced to improve the ensemble learning and establish a baiting model for river crab culture. The particle swarm optimization algorithm was introduced to improve the ensemble learning, and the bait quantity prediction model was established to realize the accurate prediction of the bait quantity of river crab aquaculture. The results of practical application tests showed that the average absolute error （MAE） of this model was 0.349 71 g, the root mean square error （RMSE）was 0.491 14 g, and the coefficient of determination （R2） of key performance reached 0.903 58.