The data in the database of ideological and political education system often come from multiple channels, including school administrative offices, online learning systems, mobile devices, etc., which have different formats, structures and quality standards. In order to unify the format structure and quality standards of the data and ensure the accuracy and effectiveness of the detection, an outlier detection method for ideological and political education system database based on deep learning is proposed. Mining ideological and political education system database anomalies, collecting and organizing data in the ideological and political education system database, and extracting useful features from the raw data as a prerequisite for outlier detection. Based on deep learning, construct the ideological and political education system database outlier detection model, and design the network structure of input layer, hidden layer and output layer according to the characteristics of the abnormal data in the ideological and political education system database to ensure the integrity of outlier detection. Generate the conditioned dissimilarity of abnormal value detection of ideological and political education system database, assess the degree of similarity or difference between abnormal points and normal data, quantify the differences between abnormal values, and avoid the problem of detection errors. The experimental results show that the method has better outlier detection performance for ideological and political education system database.