Abstract: The construction of "smart prosecution" has made great progress in recent years. However, most of the achievements are mainly in the field of prosecution informatization and data infrastructure development, with very few researches and very limited attention on decision support for prosecution. To fill this gap, based on factor that the core task in prosecution is to "prosecute criminal offenses" and prosecutors "decide which charge to prosecute according to the basic case information", this paper uses text mining techniques to establish a text classification-based decision support system for prosecution. This system mainly consists of text pre-processing, feature extraction, classification and other processes while it inputs the cases description text and automatically outputs the corresponding prosecution charges. Experimental results show that this system can achieve high accuracy with various classification models, different number of features input and different text vector representation methods. It not only achieves effective and high-precision prosecution decision support, but also improves the efficiency of case file classification management. Our research work in this paper is a pioneering attempt in the field of big data mining-assisted prosecution decision making and a concrete practice of "promoting the intelligence of prosecution". Our findings enrich the field of research and provide relevant data and conclusions that can be used as a baseline for future application and practice in this field.

Key words: text mining, text classification, decision support in prosecution, cases files classification and management