论文部分内容阅读
通过389例犯罪量刑案例的整理,筛选出14个量刑指标,采用k-means方法对一审刑量进行了分级,以此为基础构建了Logistic量刑模型。利用量刑指标间相关性,对模型进行了优化,得到犯罪危害、赔偿、自首等指标精简的量刑模型,并就模型的效力进行了评估。案例回代表明模型量刑可靠率达89.3%,有一定的实用价值,为大数据背景下的司法改革中的数理基础研究提供有益探索。
Through the finishing of 389 criminal sentencing cases, 14 sentencing indicators were screened, and the k-means method was used to grade the first-instance sentencing penalties. Based on this, a Logistic sentencing model was constructed. The model was optimized by using the correlation between the sentencing indicators, and the sentencing model was simplified, which included the crime hazard, compensation and voluntary surrender, and the effectiveness of the model was evaluated. The case back on behalf of the model sentencing reliability rate of 89.3%, has a certain practical value, provide a useful exploration for the mathematical basic research in judicial reform in the context of big data.