Statistical Analysis of Factors Influencing Agricultural Land Use Effectiveness in Russia

Agriculture is an important part of the Russian GDP and the employment sector for many Russians. Horticulture is the largest agricultural sector in Russia. Crop production has recently undergone a renewal thanks to the intensification and application of modern innovative field control technologies, new yield forecasting models and computer vision systems. There are several large Russian projects for the development of computer vision for analytical purposes in agriculture, but at this stage they are not completed. Thus, there are no common databases in Russia, as in many other countries, with satellite images and tabular information about the fields of all regions. According to that fact, there is a restriction for analytics, on the basis of which optimal decisions could be made. It is a problem for the development of the industry. The purpose of this study was to analyze the effects of various variables on crops and yield prediction. The research dataset includes the yield in rubles per hectare as the resulting variable and several groups of agronomic, environmental, climatic, soil, and socio-economic features as independent factors. The sample consists of 100 observations, whereas 75 of them form the training set and the other 25 are included in the test set. The used statistical methods include truncated regression and neural networks. One of the features of this article is the use of satellite imagery as features in neural networks models. The main hypothesis of the study is related to that fact. Field images can be an important source of data for agricultural statistical analysis and yield forecasting. The results of the analysis can be used to enhance the efficiency of the agricultural system, increase productivity, reduce the cost of production, minimize starvation problem in some regions of the world and improve the environment in rural areas. The research prospects include training and testing models on a larger sample of both tabular data and satellite images. There are also opportunities to use new exogenous variables.