Research and Application of Random Forest and ARIMA Based Modelling

In this paper, more efficient data analysis methods and advanced monitoring equipment will be explored to enhance the accuracy and reliability of the early warning system. The data are preprocessed for missing values and outliers, and the processed data are subjected to feature extraction to compute their statistical and frequency domain features. A random forest classification model containing 1000 decision trees is trained using the extracted features. By inputting electromagnetic radiation and acoustic emission data into the model for classification determination during a specific time period, high precision identification of interference signals is successfully achieved, and the accuracy of the random forest model reaches 99.99%. Using the sliding window technique, the sliding mean, sliding standard deviation and the energy of the signal were calculated for each window. Then an ARIMA model was established to fit the data to better observe the trend of the data.