Application of Logistic Model in International Oil Price Forecasting and Analysis of Sudden Factors

The high volatility and low stability of the international crude oil market make it extremely challenging to analyze and predict oil prices, but the economic returns translated from the transformation of research results are extremely lucrative and attractive to institutional investors. In this paper, we first optimize the traditional logistic regression algorithm for the row and column dimensions of the dataset, so that its convergence speed and gradient decrease, and then introduce Lasso regularization to do further dimensionality reduction. The improved logistic regression algorithm is then used as the basis for constructing an international oil price prediction model, and the prediction performance of the constructed model is compared with that of other five logistic regression models on the dataset, and at the same time, the prediction of international crude oil price is carried out for two kinds of contingencies, namely, natural disasters, and over-expectation of important data. It is found that the AUC of the model in this paper is 0.034 higher than that of the traditional logistic regression model, and the probability that the international crude oil price will keep fluctuating between $0 and $4 in the next 24 hours is the highest when natural disasters occur, with an average probability of 0.58666, while the international crude oil price will keep fluctuating between $4 and $10 in the next 24 hours when the economic data, monetary data, crude oil production and sales, and stockpiles do not meet the expectations in the best way. Maintain the greatest probability of fluctuation between 4 and 10 dollars, with an average probability of 0.55345.This study provides a new basis for trading decision-making for professional institutional investors and algorithmic trading practitioners, which helps to improve the level of trading decision-making.