在这个知识与经济高速发展的时代，社会对人才的要求与日俱增，我国研究生群体激增，每名研究生面临的竞争压力也越来越大，这严重影响着研究生的心理健康状况。在研究生群体中，由于患有心理疾病，进而导致无法正常完成学业，影响到自身的发展，这样的现象层出不穷。然而，目前我国针对该群体心理健康状况的关注较为薄弱，因此，针对现阶段研究生的心理健康状况的研究分析很有必要。 本文通过对国际心理健康量表SCL-90的研究，结合我国学生群体的现状及特点，通过分析影响我国研究生心理健康的原因，对其中的评判因素进行扩展，并进行相应数据采集，采用神经网络模型对研究生心理健康状况进行预测分析。本文主要做了以下几方面的工作： (1) 针对我国学生的具体情况，在SCL-90量表中9个因子的基础上，加入户口类型、家庭成分以及是否为独生子女3个因子，作为研究生心理健康状况模糊综合评判模型的因素集，力求评判模型更适用于我国研究生群体。 (2) 对某市部分高校研究生群体进行数据采集，以本文所建立的评判模型作为依据，将影响研究生心理健康状况的主要因素作为样本输入，采用BP神经网络建立研究生心理健康状况预测模型，利用其自学习功能对网络进行训练，得到各个因素与心理健康状况的映射关系，并且对传统BP算法进行优化，得到的神经网络能更好地达到所预期的效果，说明BP神经网络具有可行性。 (3) 对研究生心理健康状况预测模型的性能进行评价，结果显示预测值与实际值较为接近；并将所建模型与国际UPI大学生人格测试进行实际应用对比，结果表明该模型能够较好得预测我国研究生的心理健康状况，在一定程度上为我国高校心理教育工作提供了有价值的研究。

With the high speed development of knowledge and economy, the requirement of the society for talents is growing, the amount of graduate student is increasing rapidly, the competition pressure of each graduate students is getting more and more, which seriously affects the postgraduates' mental health. In the graduate student population, because of suffering from mental illness, which leads to be unable to complete their studies and affects their own development, this phenomenon is endless. However, our country's concern about the psychological health of the group is relatively weak at present. Therefore, it is necessary to study the mental health status of graduate students at this stage. This paper expands evaluation factor and carries on the judgment of graduate student mental health condition by studying the traditional mental health scale researchand combinating our country student population's present situation and the characteristic. Through analyzing the main influencing factors of mental health of graduate students in our country, and collect corresponding data and the neural network model was used to predict the mental health status of graduate students. This paper mainly has completed the following several aspects of the work： (1) Basing on in the existing measuring methods SCL-90 table of nine factors add the account type, family composition and whether only child for a total of 12 factors as related factors, establish the factors set of the mental health status of students of fuzzy comprehensive evaluation model , using iterative method to calculate the membership degree, the evaluation model are more applicable in our research group. (2) Make the main influence factors of students' mental health status as input sample, establish the optimized BP neural network of postgraduates' mental health prediction model, the corresponding fuzzy comprehensive evaluation of the results are the output sample, using neural network self-learning function of the network learns, and get the mapping relationship between the various factors and their psychological health status which is able to predict the mental health of graduate students. (3) Collect the data of one university graduate students, through the self-learning of sample data and the model building by the neural network has a good prediction result, and take the practical application for several students, and compare with the UPI college students personalitytest. The results show that the predicted model can predict the Chinese graduate students mental health status better.