Analyzing Illegal Psychostimulant Trafficking Networks Using Noisy and Sparse Data
Analyzing Illegal Psychostimulant Trafficking Networks Using Noisy and Sparse Data
复制标题
使用嘈杂和稀疏的数据分析非法精神兴奋剂贩运网络
DOI：
10.1080/24725854.2023.2254357
复制
发表时间：
2023
影响因子：
2.6
通讯作者：
Midgette, Greg
中科院分区：
文献类型：
--
作者：
Bjarnadottir, Margret V.;Chandra, Siddharth;He, Pengfei;Midgette, Greg
本文采用分析方法，利用毒品证据检索信息系统中的纯度调整价格数据，绘制美国的非法精神兴奋剂（可卡因和甲基苯丙胺）贩运网络。我们使用两个假设来构建网络：（i）始发地的纯度调整价格低于目的地；（ii）价格扰动从始发地传输到目的地。然后，我们采用两步分析方法：将数据聚合问题表述为优化问题，然后构建连接状态的推断网络并检查其属性。我们发现，首先，根据最佳聚合数据集创建的推断可卡因网络解释了 46% 的轶事证据，而过度聚合的数据集为 28.4%，聚合不足的数据集为 14.5%。其次，我们的网络揭示了许多现象，其中一些与已知的现象一致，而另一些则以前未观察到。为了证明我们的方法的适用性，我们将可卡因数据分析结果与甲基苯丙胺数据的并行分析进行了比较。这些结果同样与先前的知识相符，但也提出了新的见解。我们的研究结果表明，与次优聚合数据相比，最佳聚合数据集可以更准确地描述非法药物网络。
This article applies analytical approaches to map illegal psychostimulant (cocaine and methamphetamine) trafficking networks in the US using purity-adjusted price data from the System to Retrieve Information from Drug Evidence. We use two assumptions to build the network: (i) the purity-adjusted price is lower at the origin than at the destination and (ii) price perturbations are transmitted from origin to destination. We then adopt a two-step analytical approach: we formulate the data aggregation problem as an optimization problem, then construct an inferred network of connected states and examine its properties.We find, first, that the inferred cocaine network created from the optimally aggregated dataset explains 46% of the anecdotal evidence, compared with 28.4% for an over-aggregated and 14.5% for an under-aggregated dataset. Second, our network reveals a number of phenomena, some aligning with what is known and some previously unobserved. To demonstrate the applicability of our method, we compare our cocaine data analysis results with parallel analysis of methamphetamine data. These results likewise align with prior knowledge, but also present new insights. Our findings show that an optimally aggregated dataset can provide a more accurate picture of an illicit drug network than can suboptimally aggregated data.