连环作案嫌疑人“地理特征”估计的统计模型

A Statistical Model for Estimating the Geographic Profiling of a Serial Killer

在不同地点发生的系列谋杀案对社会的危害极大,如何确定其犯罪嫌疑人的住所是破案的关键。对犯罪嫌疑人住所的估计,给出了3种方法,即质心法、距离法和概率统计法,通过比对论证和案例检验表明,距离法更具优势;进一步展开分析,运用二维正态理论实施置信区域的估计,能根据警方所期望的可靠度和实际警力划定搜索区域,对侦破和决策具有实际指导作用;为控制精度,给出了调整方差估计的方法,构造出加权平均公式,在不降低可信度的前提下合理缩小搜索范围,为提高侦破效率提供了理论依据和有效方法。

It causes a great harm to the community that a serial murder has happened in different spots.For instance,the famous Peter Sutcliffe event is one case of such kind.A statistical model is presented providing three kinds of methods to estimate the location of the suspect＇s residence： centroid method,the distance method and probability statistics method.A conclusion is reached that the distance method has more advantages through comparison,demonstration and case testing.Further,by carrying out the all round analysis and using the theory of two-dimensional normal distribution to develop an estimation of confidence region,the model helps determine a search region,according to the reliability the police expected and the actual police forces,which is practical in solving the crime problem.Moreover,the model develops a method to narrow down the search region reasonable without reducing any reliability,by adjusting the variance estimation and constructing a weighted average formula,which provides both theoretical basis and applied methods for improving the detection efficiently.