| 摘要: |
| 针对绿色农产品消费行为具有多变量非线性相互作用的特点，传统统计方法难以准确预测消费行为的问题，提出基于改进果蝇算法优化的广义回归神经网络消费行为预测模型。首先针对果蝇群搜索不均匀所导致果蝇飞行单一的问题，提出一种均匀的果蝇群搜索机制即扇形果蝇优化算法加快搜索能力和效率；其次针对广义回归神经网络的平滑因子易受人为选择的影响，提出改进果蝇算法优化广义回归神经网络参数，实现参数的自动化选择，提高模型的预测能力。运用提出的模型对绿色农产品消费行为预测。结果表明:相较于广义回归神经网络，遗传算法优化广义回归神经网络、粒子群算法优化广义回归神经网络、果蝇算法优化广义回归神经网络和改进果蝇算法优化广义回归神经网络模型在均方根误差指标上分别下降4.45%、1.89%、4.54%和5.03%，表明遗传算法、粒子群算法、果蝇算法和改进果蝇算法能够优化广义回归神经网络模型的平滑因子，提高模型的预测精度。从平均绝对误差、均方误差、均方根误差3个评价指标看，改进果蝇算法优化广义回归神经网络模型比其他6个单一预测模型具有更高预测精度。结果证明了改进果蝇算法优化广义回归神经网络模型在绿色农产品消费行为预测的有效性，可以为供应者调整生产计划和企业营销活动提供决策支撑。 |
| 关键词: 改进果蝇算法 广义回归神经网络 消费行为 预测 绿色农产品 |
| DOI：10.11841/j.issn.1007-4333.2023.12.19 |
| 投稿时间：2023-04-27 |
| 基金项目:国家重点研发计划(2021YFC3340501) |
|
| Optimizing the generalized regression neural network based on improved fruit fly algorithm for green agricultural products consumer behavior prediction |
|
LI Bo1,2,LIAO Mengjie1,3,ZHANG Jian1,3*
|
| (1.School of Economics and Management, Beijing Information Science and Technology University, Beijing 100192, China;2.School of Economics and Management, Beijing University of Technology, Beijing 100124, China;3.Beijing Key Lab of Green Development Decision Based on Big Data, Beijing 100192, China) |
| Abstract: |
| To address the problem that the consumption behavior of green agricultural products has multivariate nonlinear interaction characteristics, and it is difficult for traditional statistical methods to accurately predict consumption behavior, a generalized regression neural network model based on improved fruit fly algorithm for consumer behavior prediction is proposed. Firstly, to address the problem of single fruit fly flight caused by uneven fruit fly swarm search, the study proposes a uniform fruit fly swarm search mechanism, i. e. , sector fruit fly optimization algorithm to speed up the search ability and efficiency. Secondly, to address the problem of the smoothing factor of generalized regression neural network is susceptible to human selection, an improved fruit fly algorithm is proposed to optimize the selection of generalized regression neural network parameters to automate the selection of parameters and improve the prediction ability of the model. The proposed model is applied to predict the consumption behavior of green agricultural products. The results showed that: Compared with the generalized regression neural network model, the genetic algorithm optimized the generalized regression neural network, and the particle swarm algorithm optimized the generalized regression neural network, fruit fly algorithm optimized the generalized regression neural network and improved fruit fly algorithm optimized the generalized regression neural network models decrease 4. 45%, 1. 89%, 4. 54% and 5. 03%, respectively. Above results indicated that the genetic algorithm, particle swarm algorithm, fruit fly algorithm and improved fruit fly algorithm could optimize the smoothing factor of the generalized regression neural network model and improve the prediction accuracy of the model. From the three evaluation indexes of mean absolute error, mean square error, and root mean square error, the improved fruit fly algorithm optimized the generalized regression neural network model displaied higher prediction accuracy than those of the other six single prediction models. The results demonstrate the effectiveness of the improved fruit fly algorithm optimized the generalized regression neural network model in predicting green agricultural products consumption behavior, which can provide decision support for suppliers to adjust production plans and corporate marketing activities. |
| Key words: improved fruit fly algorithm generalized regression neural network consumption behavior prediction green agricultural products |
