Research on the Application of Time Series ARIMA Model and Neural Network Model

This article uses ARIMA model and BP (Back Propagation) neural network model, which is a nonlinear system with the most significant feature of high adaptability. This algorithm takes n samples of the research object as the input layer nodes of the neural network, takes the expected results as the corresponding output nodes, calculates weights and thresholds, and calculates the error values of the actual and predicted results. K-Means is a commonly used key method. Firstly, we need to randomly determine K as the initial point centroid. Then, each point in the dataset is assigned to a cluster and will be assigned a cluster corresponding to its centroid. The main idea of SVM is to find a hyperplane in the feature space, thereby separating as many data points of different categories as possible and maximizing the distance from the hyperplane. For linearly non separable data, it is necessary to select an appropriate kernel function to map the data into a high-dimensional space, so that the data can be linearly separable within the high-dimensional space. We use numerical optimization methods such as sequence minimum optimization (SMO) to solve the objective function and find the hyperplane parameters and thresholds that maximize the margin. AdaBoost (Adaptive Boosting) is an ensemble learning algorithm. It iteratively trains a series of weak classifiers and adjusts sample weights based on their performance to obtain higher weights for misclassified samples, thereby making the next weak classifier more focused on sampling these misclassified samples. Regarding the selection of weak classifiers, a weak classifier with poor performance will be selected in each iteration, in order to pay more attention to these misclassified samples in the next iteration and improve overall performance.