Using Elastic-Net for High Dimensional Time Variables Selection of Autoregressive Model Series of Caenorhabditis Elegans Motion

The process of selecting statistical variables that contain information related to the effect on the dependent variable has a fundamental role in accurate statistical modeling. In time series and when there are a large number of an autoregressive (AR) variables in the model, selecting the most effective AR variables among a large set of autoregressive variables (high dimensional) with prior time lags is important for more accurate results. The elastic net method is one of the methods used for selecting the best model and conducting a joint estimation of the linear models, which contributes to the selection of the actually influencing variables and ignoring others among a very large set (high dimensional) of autoregressive variables. In this study, the elastic network method will be used to select and estimate the autoregressive parameters in the time series model. Caenorhabditis elegans (CE) will use time series data for the movement of Caenorhabditis elegans, represented by tangent angles of the wave motion. The univariate time-series model of CE movement was selected via the elastic network method and the hybrid (Elastic-AR) autoregressive model after multi-processes of selecting autoregressive variables. According to the results, the selected parameters in the AR model matched the Elastic-AR hybrid model, with clear superiority in the results of the hybrid method and with high accuracy. Therefore, it is possible to conclude the possibility of using the proposed hybrid method to obtain the best model for the high-dimensional time series dataset with the least number and the most influential of variables, which reduces effort and costs and increases the accuracy of these models..