With the expansion of universities, university students who are in their youth will inevitably fall into psychological states such as panic and anxiety in the face of increasingly fierce employment competition and severe employment situation. This paper uses the association rule Apripri algorithm to construct a correlation analysis model of university students' psychology health and employment based on data mining. By excavating the psychological problems of university students for specific analysis, better countermeasures can be drawn. This article has a lot of support based on the results of the research to date. Data mining is a means of discovering regularities from a large amount of data by analyzing individual data (Witten et al. 2018). This is the decision support process (Baghernia et al. 2016). Data mining is divided into directed data mining and undirected data mining (Zhao et al. 2018). The learning means of neural network is mainly reflected in the change of weights (Witten et al. 2019). Decision trees are rules for building classifications based on various utilities of target variables (Song et al.2019). Genetic algorithms are often used to optimize neural networks (Bala et al. 2017). Rough set means is a new mathematical tool for dealing with ambiguous, incorrect and incomplete problems (Rehman et al. 2019). Fuzzy set means describes fuzzy attributes according to membership degree (Hastie et al. 2019). Association rules reflect the interdependence or association between things (Keogh et al. 2018). Data mining is a new business information processing technology (Shaferet al. 2017). Its goal is to discover implicit and meaningful knowledge from databases (Khalil et al. 2018). Employment psychology refers to various psychological phenomena that occur in the process of people considering employment issues, preparing for employment and employment (Chen et al. 2017). Emotional problems are noteworthy problems in the psychology health of university students (Li et al. 2021). Building a sound psychology health education system for university students is the basic means to prevent psychological crisis (Alcala et al. 2017). In the process of employment, university students not only show a stable and confident psychological state, but also show changing and complex feelings (Liao et al. 2018).The second part of the paper explains the relevant theories, and introduces the mining process and classification algorithm in data mining. The third part explains the correlation between college students' mental health and employment. The fourth part explains the method used, and the result is good. The above literature provides a good reference for the technical route of this paper and provides a good theoretical basis. Firstly, we choose physiological signals and text signals as information sources. We can use the mature convolution neural network to extract features for psychological evaluation, and introduce random forest algorithm to determine the degree of mental health. According to the two special properties of psychological state, the weighted fuzzy reasoning theory is introduced to improve the analysis method. Through evaluation, colleges and universities can get psychological early warning, know students' mental health status as early as possible, correct bad academic psychological status in time, and reduce the negative effects brought by employment.