浓度算法.docxVIP

Model 1 浓度算法： （离散模型，确定多个可能解，极端情况，样本数量的变化不影响算法的可靠度，无中心概念） Overview （离散模型，确定多个可能解，极端情况，样本数量的变化不影响算法的可靠度） 我们的模型一建立了一个离散模型。该模型将杀人犯的运动（即从一个案发地到另一个案发地的移动）假想为猎物浓度驱动的，同时猎物浓度也会在案件的连续发生过程中而动态改变。 我们构造了一个算法(concentration-calculator)来通过作案区域内部的正常浓度分布、所有已发案件的分布和时间来确定当前的作案区域的浓度；同时我们构造了一种基于离散化的最速下降法的搜送策略(points-generator)，使杀人犯根据当前的作案区域中个小正方形的猎物浓度来决定下一步怎样走。 事实上，该模型中参加每一部迭代的点都有多个，其中每一个迭代点又会各自产生自己衍生出的多个迭代点。 The Concentration Driven Model (CDM) is a discrete model. In this model, we suppose that the murderer’s moving from one action space to another depends on preys’ distribution through the hunting area. Meanwhile, as the crimes being committed sequentially, the distribution will change. We develop an algorithm (we call it concentration calculator) to calculate the current preys’ distribution using the happening time and the location of the past crimes scenes. Then, we develop an algorithm (we call it points-generator) to search the next possible crime locations. The points-generator is a ‘discrete ‘steepest descent method. Frame of the Model(改) Definition Concentration calculator Change of the prey’s distribution（缺点，当作案范围大时，猎物浓度不会变化） According to our assumption above, each action space own a property of prey concentration. So we investigate the transfer of preys among the many action spaces to estimate the change of the prey distribution. 两种猎物浓度变化的原因及算法： Causes for transfer : ? 瞬时变化原因：刚刚发生了杀人案（对这种变化的一种假想，上下班的人绕路走） Causes for instant transfer :a crime has just been committed 当一个小正方形区域中刚刚发生了杀人案，这个区域的猎物浓度会发生瞬时的变化，此时我们将案发的小正方区域设为Ci,j;此时设所有猎物会向邻近的8个区域转移，我们将与Ci,j邻近的8个区域分为A、B两类，如下图所示: According to our assumptions, when a crime has just been committed in an action space, all the preys in the action space will soon run away to other action spaces. This assumption is reasonable. In fact, people who will emerge in a certain place always for the reason that he will pass through the place on the way to work or some places else in his daily life, and he will go round if an crime has ju