基于序贯指示模拟的信息预测方法

An Information Prediction Method Using SISIM

信息预测在医药、地质、军事等众多工程应用领域中发挥着重要作用。在模拟过程中仅利用少量的已知数据预测未知信息非常困难。为了获得更好的模拟效果,在估计或预测未知区域的过程中采用内插和外插的方法。作为一种非参数的估计算法,指示克里格方法虽然能够在给定周围条件数据的情况下估计任何位置的条件累积分布函数,但并不能保证结果分布的有效性。序贯指示模拟(Sequential indicator simulation,SISIM)综合了指示克里格方法和序贯方法在模拟非参数连续和离散分布函数上各自的优点。该方法根据一系列自定义阈值确定非参数累积分布函数,模拟过程中利用底部和顶部外插函数给出分布函数的形状。试验结果表明模拟结果具有与真实数据相似的分布特征,证明了该方法的实用性。

Information prediction plays an important role in many fields such as medical,geological and military fields.However,it is quite difficult to predict the unknown information only by some sparse known data in the process of simulation.Therefore,some interpolation or extrapolation methods are used to estimate or predict the unknown region for better simulated results.Indicator kriging is a non-parametric estimation algorithm used to estimate the conditional cumulative distribution function at any location given neighboring conditioning data,which cannot guarantee that the resulting distributions are valid.Sequential indicator simulation（SISIM） combines the indicator kriging with the sequential method to simulate non-parametric continuous or categorical distributions.In SISIM,a non-parametric cumulative distribution function is determined from a set of threshold values.The lower tail extrapolation function and the upper tail extrapolation function provide the shapes of the distribution functions during the simulation of SISIM.Experimental results show that the simulated results have the similar characteristics of distributed values with the original one,demonstrating that the proposed method is practical.