Anomaly Detection of Time Series Based on SSA-LASSO

In order to improve the accuracy of time series anomaly detection and classification, aiming at the problem of large feature dimensions of high-dimensional data, on the basis of studying the traditional feature selection algorithm of Lasso (Least Absolute Shrinkage and Selection Operator), combining Sparrow Search Algorithm (SSA) to optimize the Lasso feature selection algorithm, by optimizing the regularization parameters of Lasso and combining the strength of maximizing the penalty weight, improves the stability of the traditional LASSO algorithm in feature selection and effectively improves the robustness of the model. When LASSO performs feature selection, it retains the feature items with high coefficient correlation. At the same time, it uses the SSA algorithm to optimize LASSO feature selection to avoid the overfitting problem caused by too many features, so as to achieve the optimization ability of feature selection and the accuracy of classification. The Python experimental results have verified that this method can effectively improve the accuracy of classification.