调解员特征与调解成功率相关关系研究——基于一种可解释机器学习方法

本文对H省“人民法院调解平台”1542名调解员办理的21万件调解案件,采用融合XGBoost算法与SHAP解释框架的机器学习方法,以可视化的方式对调解员身份、学历、性别、经验四个特征与调解成功率的相关关系进行分析。结果表明:在该模型中,调解员的身份对调解成功率的影响最大,且大多数情况下当调解员身份为行政调解员或法官时,调解员身份会对调解成功率起到正向贡献作用;调解员经验的重要性仅次于调解员身份,但当经验值达到一定数值后,调解成功率不再受调解员经验的显著影响;相较于调解员身份和经验,学历对调解成功率的影响较小,随着学历的提高,其对调解成功率逐渐产生正向贡献,但当学历提高到硕士、博士时,其贡献作用明显减弱甚至出现负值;调解员性别对调解成功率的影响最小,且男性的影响小于女性,调解成功率更容易受到女性调解员的影响。

In this paper,210,000 mediation cases handled by 1,542 mediators in“People’s Court Mediation Platform”in H province,the machine learning method combining XGBoost algorithm and SHAP interpretation framework is used to visually analyze the correlation between the four characteristics of mediators,such as status,education,gender and experience,and the mediation success rate.The results show that in this model,the mediator’s status has the greatest influence on the mediation success rate,and in most cases,when the mediator’s status is an administrative mediator or a judge,the mediator’s status will contribute positively to the mediation success rate.The importance of experience is second only to mediator’s status,but when the experience value reaches a certain value,the success rate of mediation is no longer significantly affected by experience;Compared with the mediator’s status and experience,the educational background has little influence on the success rate of mediation,and with the improvement of educational background,it gradually makes a positive contribution to the success rate of mediation,but when the educational background is increased to master’s degree and doctoral degree,its contribution is obviously weakened or even negative;Gender has the least influence on the mediation success rate,and the influence of men is smaller than that of women.The mediation success rate is more easily influenced by female mediators.