Pretive Role and Recommendations of Time Series Data Based on REF-RF

The paper mainly focuses on the analysis and prediction of time series data. One of the main characteristics of time series data is that it usually contains trends and patterns, which are relatively hard to measure and predict accurately. Therefore, we set up a reasonable and accurate model based on a specific dataset. The quantitative model predicts outcomes under differences in key time series indicators. Missing value imputation, outliers handling, and normalization of data are performed through data preprocessing using Box Plots, BP Neural Network Algorithms, and Dimensionless Quantification before trying to solve the problem. The impact of various factors on time series data is modeled in the first part of this study. After that, dimensionality reduction for each influencing factor was done using the RFE-RF algorithm. Then, Entropy-TOPSIS should be run to express the size of impact. In this step, consider the weight of each indicator as the standard score in order to get an evaluation with respect to the importance of various factors due to the time series data. Finally, Plotting the data by making a line chart of the level difference of key indicators against time using Python. We applied the model to additional datasets to assess its accuracy. Subsequently, to check and validate this model, we implemented a Decision Tree and afterward applied regression to further fine-tune the model by its prediction.