An Optimization Method for Sentencing Prediction Based on the Nonlinear Recursive Identification Theory and Clustering Algorithm *

Artificial intelligence assisted sentencing is a pivotal component of the Smart Court Project in China. This paper proposes an optimization method (Law K-means) for predicting sentences, based on the K-means algorithm and the nonlinear recursive identification theory established in [1]. This method incorporates sentencing logic into the clustering algorithm then constructs 21 sentencing intervals. Sentencing intervals are constructed on the basis of actual sentences, which is more practical to judges as opposed to artificial divides. Compared to K-means clustering algorithms, the proposed method shows a better predictive performance of sentencing prediction.