Abstract:
This article describes an algorithm for applying an intelligent analysis model to detect anomalies in statistical observation data for educational organizations. The definition of an anomaly is given, typical anomalies that may be contained in statistical reporting data are analyzed. The classification of anomaly detection techniques is given depending on the level of markup of the training sample, and possible ways of marking up data to present the results of the anomaly search are analyzed. The analysis and description of the process of collecting and processing statistical data of educational organizations in the Scientific and Technical Center of RTU MIREA is carried out. The weaknesses of the data collection process are analyzed, which can be strengthened by applying intelligent analysis to search for anomalies in the data. The analysis and mathematical description of the format and features of the received and stored statistical data is carried out. An algorithm has been developed for preparing data for training an intelligent analysis model, taking into account their specifics, as well as the subsequent application of the trained model to detect anomalies in the data under consideration. The algorithm was tested on real data using the autoencoder neural network model.
Keywords:anomaly detection, statistical data, data mining, autoencoder.
