Method for predicting air quality grade by integrating sequence pattern mining and cost sensitive learning

Abstract

The invention relates to a method for predicting the air quality grade by integrating sequence pattern mining and cost sensitive learning, which comprises the steps of firstly mining a sequence pattern from historical sequence data of the air quality grade, and building a sequence pattern tree; then extracting features from air quality and meteorological historical data, and training a cost sensitive prediction model based on a cost sensitive learning technology; and finally integrating the sequence pattern mode and the cost sensitive prediction model for final air quality grade prediction. The method takes a changing pattern of the air quality grade and the unbalanced prediction error cost into consideration on the basis of the existing air quality grade prediction method based on machinelearning, and can effectively improve the prediction performance of the model.

Description

technical field [0001] The invention relates to the technical field of machine learning, in particular to an air quality level prediction method that combines sequential pattern mining and cost-sensitive learning. Background technique [0002] Air pollution is a very serious urban problem at present. Monitoring and forecasting air quality level is of great significance for pollution assessment, pollution control and harm reduction. At present, the air quality level is monitored in real time through air quality monitoring stations, and the prediction of air quality level needs to be realized by designing a reasonable calculation model. [0003] The existing air quality level prediction methods mainly fall into the following two categories: [0004] (1) Atmospheric pollutant diffusion calculation model: It is an empirical method, which calculates the air quality level at different locations after a period of time through parameters such as pollutant concentration, wind direct...

AI Technical Summary

Problems solved by technology

However, such methods also have the following problems: First, most existing methods train discriminative models (such as decision trees, support vector machines, etc.)

In fact, the classification error of the air quality level cannot be treated uniformly (for example, assuming that the real air quality level is 2, the impact of mispredicting it as 5 is obviously greater than that of mispredicting it as 3)

Images