The increasing number of graduates from private colleges and universities creates a great challenge to analyze the factors influencing the employment of graduates and to improve the quality. Based on the structural equation SEM model and combining the data of graduates of private colleges and universities in a city, this paper integrates and selects the factors influencing the employment quality of graduates of private colleges and universities. An evaluation model for the employment quality of graduated students has been established through the LSTM network to further study the degree of influence of each influential factor on graduates’ employment quality. In order to deal with the bidirectional dependence of time series data while considering the inputs before and after the current moment, this paper adds the reverse LSTM layer on top of the LSTM network to form a BiLSTM network, which better captures the correlation between the influencing factors. In the research process, it was found that, except for the first selected factor of graduates’ destination, which needs to be eliminated, the other six influencing factors passed the reliability and validity test and could make the model have a good fitting-out effect. In the algorithm analysis, it was found that the running accuracy of the BiLSTM network could reach 97.01% at the maximum. In comparison, the running accuracy of the LSTM network could only get 91.03% at the maximum.BiLSTM network is more outstanding than the LSTM network both in accuracy and stability.