Novel Abnormal Data Mining and Detection Algorithm for Intelligent Systems

In intelligent systems, as the amount of data increases, how to analyse data and extract abnormal information is an important task. Based on this, in order to improve the detection efficiency and accuracy of abnormal data in intelligent systems, this study proposes a new and efficient abnormal data mining and detection algorithm. In the initial modelling process, this study introduced the minimum modification cost principle based on cross entropy for modelling. This algorithm optimizes the time series data cleaning process, thereby removing noise data while maintaining the integrity of the data sequence. Then, the extreme gradient boosting algorithm is combined with the Gaussian kernel function to achieve data dimension reduction. This operation can reduce data dimensions and feature redundancy. Finally, the overall anomaly detection operation by integrating the improved Ada boost algorithm with the sampling balance technology can improve the accuracy of the classifier while solving the problem of unbalanced data. The experimental results show that the proposed method is superior to the traditional algorithms in comprehensive indicators.