This paper focuses on space-time nonlinear intelligent modeling for lattice data.Lattice data refers to attributes attached to fixed

regular or irregular

polygonal regions such as districts or census zones in two-dimensional space.Lattice data space-time analysis is aiming at detecting

modeling and predicting space-time patterns or trends of lattice attributes changed with time while spatial topological structures are simultaneously kept invariable.From the perspective of space

lattice objects have two different scale spatial properties influencing lattice data modeling: global dependence and local fluctuation.Global spatial dependence or autocorrelation quantifies the correlation of the same attribute at different spatial locations

and local spatial fluctuation or rough

coexisted with global dependence

is represented in the form of local spatial clustering of similar values or local spatial outliers.To consider simultaneously the effects of two properties above

local neural networks（NN） model is studied for space-time nonlinear autoregressive modeling.The main research contents include:（1） To reduce influence of spatial fluctuation on prediction accuracy of NN

all regions are partitioned into several subareas by an improved k-means algorithm.（2） Different partition schemes are evaluated and compared according to three essential criteria including dependence

continuity

fluctuation.Dependence means that an optimal partition must guarantee that there is real and significant spatial dependence among regions in a subarea because the results of output layer nodes in a NN model depending on the interactions of input layer nodes through hidden layers nodes.Spatial autocorrelation of a subarea can be measured by global Moran’s I and its significance test can be done based on z-score of Moran’s I.Continuity means that only neighboring regions can be grouped into a subarea

and this criterion is fused into the modified k-means algorithm.When the algorithm judges one region which subarea it belongs to

not only should the distance be considered to the centroid of a subarea but also the common borders between this region and the regions in a subarea.As to fluctuation

although it is impossible to make each subarea have complete spatial stability through partitioning

the less fluctuation means the better predicting results of NN model.For a subarea

standard deviation between local Moran’s I of all regions in the subarea and global Moran’s I of the subarea is regarded as an evaluation index to the fluctuation of the subarea.（3） Each multi-layer perceptrons（MLPs） network is used respectively in modeling and predicting for each subarea.The output nodes are the predicting values at time t of an attribute for all regions in a subarea.The input nodes are observations before time t of the same attribute of both regions in the subarea and regions neighboring to the subarea and the latter is called boundary effect.Finally

as a case study

all local models of all the subareas are trained

tested and compared with a single global MLPs network by modeling one-step-ahead prediction of an epidemic dataset which records weekly influenza cases of 94 departments in France from the first week of 1990 to the 53th of 1992.Two performance measures

including average relative variance（ARV） and dynamic similarity rate（DSR）

indicate that local NN model based on partitioning has better predicting capability than global NN model.Several issues are still worth further study:（1） The initial subareas of partitioning are selected randomly in our research.In the further study

a reasonable approach should combine selection with spatial patterns

for instance considering the center of local cluster.（2） Partition criteria should be another issue and different types of spatial and space-time processes

such as rainfall

price waves

public data

etc

may have different objective criteria for choosing an optimal partition.（3） It may be more imperative to study feasible measures for quantifying global and local space-time dependence of lattice data and testing significance of this dependence.