An anomaly detection method based on Lasso

In many research and application fields, anomaly detection always is an important issue. In the article, a method of anomaly detection is presented which based on Lasso on the basis of variable linear regression solution of the Lasso problem. We transform the process of anomaly detection into a linear regression model, meanwhile, take the detection parameter as regression variables and establish the model of the regression variables and the dependent variable. Due to estimation of Lasso parameter own stable regression coefficient, can compress parameters of the model and reduce the number of parameters. Those characteristics accord with the requirement of stability, high-speed and simplicity which are needful for anomaly detection. Experimental results show that our method has higher detection accuracy and more rapid convergence ability under the constraints of the appropriate threshold.