Using Linear Regression for Weather Prediction

In recent years, the direction of numerical forecasts has been actively developed, based on the analysis and processing of observational data from previous periods in order to identify patterns of their change and obtain values of climatic parameters for future periods. The article deals with the application of linear regression for weather forecasting. Based on weather observations in Nizhny Novgorod, a test data set was prepared, features that have the greatest impact on the predicted value were identified – the maximum, minimum and average temperature over the past days, the maximum, minimum and average dew point values. Studies have shown that it is not necessary to consider a large number of features over a long period of time to make accurate weather forecasts - a small set of data from the previous day may suffice. The average forecast error was 1.5 degrees.