Comparing the performance of Ridge Regression and Lasso techniques for modelling daily maximum temperatures in Utraradit Province of Thailand

Daily maximum temperature is one of the important climate components for every country. There are many explanatory variables that may affect a response in a particular model, but it is not easy to identify the significance of these effects. Selection techniques are required to assess this significant issue. Using methods to shrink the number of explanatory variables in order to produce a simpler model. This study compares the performance of Ridge regression and Lasso regression in modeling temperature data for Utraradit Province, Thailand. Daily maximum temperature data from 1 January 2015 through 31 August 2021, in total 2435 days were used. The response variable is the daily maximum temperature with 32 explanatory variables. The Ridge regression model successfully selected and reduced the number of variables from 32 to only 13 variables. In addition, based on the MSE and coefficient of determination (R 2 ), Ridge regression provided better performance when compared to Lasso regression. The resulting Ridge regression model showed that the average, maximum and minimum humidity, the minimum humidity of the previous day, the average wind speed, the average and maximum wind speed of the previous day, the average, maximum and minimum solar radiation, the average, maximum and minimum solar radiation of the previous day significantly affected the daily maximum temperatures. On the other hand, Lasso regression revealed that daily maximum temperatures were significantly affected by the average solar radiation, the maximum solar radiation of the previous day and the maximum and minimum barometric pressures.