Multi-variable multi-valued regression data set integration method

Abstract

The invention discloses a multi-variable multi-valued regression data set integration method, and relates to the field of data research and prediction. The invention comprises the following steps of:S001, obtaining historical data of a target economic index and related peripheral indexes; S002, cleaning the data and integrating corresponding explanatory variables and response variables; S003, building a neural network model and debugging the network structure and parameters; S004, multiple training the model, and selecting multiple sets of models based on the expressions on the verification set, and combining the results. According to the multi-variable multi-valued regression data set integration method, against the multi-dimensional economic index prediction problem, a plurality of peripheral variables are introduced, based on a neural network algorithm, a Long Short Term Memory (LSTM) structure is adopted, and the random selection verification set is improved, a plurality of screened models are combined, so that time dimension is introduced to the algorithm based on the traditional multiple regression, and thus the relational information in the historical data is retained in the model. The prediction result is improved, and economic index prediction precision and stability are improved.

Description

technical field [0001] The invention belongs to the field of data research and forecasting, and is used to predict target economic indicators in a period of time in the future by using relevant peripheral indicators under the condition of ensuring accuracy and stability, especially involving a kind of data of multivariable and multivalued regression method of integration. Background technique [0002] In the field of economic data research, it is necessary to predict economic data such as gross national product, industrial added value and other indicators, so as to adjust relevant policies and make the local economy develop benignly. [0003] When forecasting economic indicators, it is necessary to consider the relationship between the relevant external indicators and the economic indicators, and model forecasts based on this relationship. Because economic data are directly or indirectly related to a large number of external indicators, it is difficult to describe the relat...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a data set integration method for multivariate and multivalued regression. Aiming at the problem of multidimensional economic index prediction, a plurality of peripheral variables are introduced, based on neural network algorithm, long short-term memory (LSTM) structure is adopted, and random Select the verification set method to improve, and combine multiple screened models, so that the algorithm introduces the time dimension on the basis of traditional multiple regression, thereby retaining the relationship information in the historical data in the model, improving the prediction results, and solving the current problem. Some economic indicators have insufficient forecast accuracy and stability

Images