Regression analysis of the dependence of the cherry yield from hydro-thermal factors in the conditions of multicolinearity

In the conditions of the Southern Steppe zone of Ukraine the influence of weather factors of the region on the formation of cherry yield within the limits of 2007–2019 years of researches is revealed. The correlation analysis allowed us to identify ten weather factors that have a notable (significant) and strong, both direct and inverse linear correlation with cherry yield (r = 0.68… -0.86). The model describing the influence of hydrothermal factors on cherry yield is proposed. Multicollinearity effects were detected when analyzing data using paired correlation coefficients and VIF. In this case, using the least squares method to construct and analyze the regression model is inefficient. A regression model based on the LASSO method is proposed. The LASSO method allows reliable estimation of regression parameters. On the basis of the constructed model the factors influencing the cherry yield index are analyzed and it is shown that the most significant factor is the average monthly rainfall for August, mm, then the factor is the absolute minimum relative humidity in May, %, the amount of rainfall during the flowering period, mm, the amount of the effective flowering temperatures, °C, the difference between average maximum and minimum flowering temperatures, °C, the total number of days with precipitation during flowering period, day, the average of maximum air temperatures during flowering period, the amount of active temperatures during flowering period, °C, the amount of active temperatures during the growing season (before the fruit ripening phase), °C, the hydrothermal coefficient (HTC) during flowering period. The indices of the share of influence of each factor on the total variance of the cherry yield index are determined. The share of influence of the factor is x1 – 9.90 %, factor x2 – 5.56 %, factor х3 – 1.95 %, factor x4 – 8.22 %, factor x5 – 14.83 %, factor x6 – -10.12 %, factor x7 – 12.12 %, factor x8 – -3.81%, factor x9 – 27.96 %, factor x10 – 5.54 %.