High-Dimensional Robust Prediction Models And Its Application In Inflation Rate

Indicator of a country’s economic stability.Forward-thinking inflation forecasts help central and other government departments to formulate effective monetary policies to stabilize prices and help financial institutions and investors make better investment decisions.Accurate forecast of future inflation rate plays an important role in the formulation and implementation of monetary policy.That is,the central bank can make a reasonable forecast of the future economic situation based on the inflation forecast,reduce the effect of deviation caused by the lag of monetary policy,and make macro-control more accurate and effective.There are many kinds of inflation prediction models,utoregressive,factor models,etc.are commonly used forecasting models,but outliers,noise,and random perturbation terms in the forecasting process will have a certain impact on the accuracy of forecasting,and the model assumes that some actual macroeconomic conditions need to be comprehensively considered.In this thesis,the hybrid and robust statistical modeling methods are used to improve the prediction effect of the model,and the prediction effect is analyzed and compared by calculating the mean square error and the average absolute error.The optimization of the prediction model in this The optimization of the prediction model in this The optimization of the prediction model in this The optimization of the prediction model in this thesis can be classified into two types: firstly,the least squares regression,factor model and Lasso regression are mixed with the random forest model,so as to study the influence of variable selection can be classified into two types: firstly,the least squares regression,factor model and Lasso regression are mixed with the random forest model,so as to study the influence of variable selection can be classified into two types:firstly,the least squares regression,factor model and Lasso regression are mixed with the random forest model,so as to study the influence of variable selection can be classified into two categories: firstly,the least squares regression,factor model and Lasso regression are mixed with the random forest model,so as to research the influence of variable selection and regression prediction process on the prediction effect of the model;secondly,Random forest,Lasso and factor models are robustly optimized for three models with better prediction effects.The robustness optimization methods of the three models are: random forest uses robust loss function in the prediction process;Lasso uses quantile regression for prediction,and obtains a high-dimensional quantile Lasso regression model;and The robust factor model uses the principal component estimation method to estimate the latent factor variables,and then predicts through the latent factors to obtain the principal component estimated factor model.Finally,an empirical analysis of national and Jiangxi inflation is conducted using single model,mixed model and robust model.The analysis results show that: for a single forecast model,the benchmark model and the machine learning model have their own advantages,but the random forest,Lasso and factor models have better forecasting effects;however,the robust forecast model and the mixed model have better forecasts for inflation.Accuracy has improved significantly,and the principal component estimation factor model and the high-dimensional quantile Lasso have the best prediction results.