CAJ | 학술논문

目标罪犯藏匿位置的预测是一个重要而艰巨的任务。提出了一种面向关系语境的目标罪犯藏匿位置预测方法。该方法主要通过在面向关系语境的社会网络中计算用户间的信任度得到目标罪犯的强关系子集，并根据所有用户的历史位置轨迹筛选出目标罪犯可能藏匿的候选位置子集，然后进行协同过滤计算，预测目标罪犯当前最可能藏匿的位置。这种方式融合了用户的社会关系及历史位置信息，有助于降低直接预测目标罪犯藏匿位置的盲目性和困难性。最后，利用模拟数据进行仿真实验，并对所提方法的有效性进行了验证。
목표죄범장닉위치적예측시일개중요이간거적임무。제출료일충면향관계어경적목표죄범장닉위치예측방법。해방법주요통과재면향관계어경적사회망락중계산용호간적신임도득도목표죄범적강관계자집，병근거소유용호적역사위치궤적사선출목표죄범가능장닉적후선위치자집，연후진행협동과려계산，예측목표죄범당전최가능장닉적위치。저충방식융합료용호적사회관계급역사위치신식，유조우강저직접예측목표죄범장닉위치적맹목성화곤난성。최후，이용모의수거진행방진실험，병대소제방법적유효성진행료험증。
It is an important and challenging task to predict hiding locations of target criminals. This paper proposes a relation context oriented approach to predict hiding locations of target criminals. The proposed approach first finds the minimal strong tie subset by computing trust strength between users in relation context oriented social networks, and selects the candidate location subset that target criminals may hide based on users’historic location trails, and then predicts the current most likely hiding locations of the target criminals by collaborative filtering computation. This way combines the social relationships and historic location information of users, thus it is helpful to abate the blindness and difficulties of predicting directly the hiding locations of target criminals. Finally, simulations on syn-thetic data are conducted to evaluate the effectiveness of the proposed approach.