基于SEM的元胞自动机转换规则蜂群智能挖掘

While cellular automata are increasingly used for modeling urban growth and land-use changes, the methods for identifying the driving factors' number and weight still require improvement. On the basis of analysis for driving factor of urban growth, the Structural Equation Modeling is introducted to determine the observable,latent and dependent variable of bee colony optimazation for trasition rule.Then, the bee feeding mechanism will be abstracted and model the cellular automata transition rules. Firstly, the Structural Equation Modeling is introducted to model and assess the urban growth through observed variables, latent variables and dependent variable against the bee colony factor randomly selection and heuristic function.The observed variables of the structural equation modeling are treated as bees mining space variable and weight coefficient as the constraint,which will be used to construct the bee colony intelligent rules. Using these transition rules cellular automata will simulate future urban growth. Lastly, the case of Beijing Changping will use this SEM-BCO to simulate urban growth and assess accuracy of model throuth the confusion matrix. In this paper the Structural Equation Modeling introduced to identify the driving factors number, weights and inherent mechanism will be important to select the conversion factor. The advantage of Structural Equation Model is to determine the quantitative impact of different factors on simulation results, reducing the blindness of factor selection.

元胞自动机模型在城市土地利用变化、城市扩张模拟等方面得到了很好的应用，但目前对模型转换规则中参数个数及权重的确定存在随意性。本项目将基于结构方程模型，构建和评估蜂群智能获取转换规则中的参数获取模型，研究蜂群算法觅食机制及抽象为元胞自动机转换规则的实现方法。首先针对蜂群参数的选择及蜂群启发函数确定问题，引入结构方程模型，通过确定元胞自动机城市扩张的观测变量、潜变量和因变量进行结构方程模型的构建和模型评估，以此结构方程模型观测变量作为蜂群挖掘空间变量，权重系数作为约束蜂群蜜源质量条件，构建蜂群智能规则，进行城市增长元胞自动机演化模拟。最后以北京市昌平区为研究实例进行城市增长动态模拟，通过混淆矩阵等检验模型精度。本研究可为智能化挖掘转换规则提供新思路，明确驱动因子对变化预测结果的定量影响，减少智能算法因素选择的盲目性。

项目针对元胞自动机模型转换规则中的参数个数选择及权重的确定随意性问题，以北京市昌平区为研究区，引入结构方程模型,确定该研究区城市扩张中各地类的观测变量、潜变量和因变量，构建了地类的结构方程模型，并应用偏最小二乘法进行了参数估算和模型评估。以随机采样样本点的因变量建立元胞属性特征空间，因变量估计参数作为特征向量优化条件，改进蜂群算法中概率选择作为转换规则，实现北京市昌平区土地利用变化模拟，并通过混淆矩阵进行模拟精度分析。实验结果证明：构建的研究区元胞自动机的结构方程模型，各地类的测量模型和结构模型的理论意义与实际意义吻合度较高，模型中各个因子与其指标之间拟合良好，所选择指标可有效反映测量因子；基于蜂群智能算法所构建的CA转换规则的数学模型，权重系数作为约束蜂群蜜源质量条件，可提高蜂群算法搜索效率，提高CA模拟精度，以解决复杂、高维、空间多目标与CA动态的优化问题。