Construction of a modified Cobb-Douglas model using a machine learning approach to optimize wheat productivity

Abstract

Food availability has become a tangible issue in the face of the prevailing threat of food insecurity. Wheat production is leading significant role to concerns over the country's ability to meet its food needs. In the current study, modified Cobb-Douglas regression model is developed using machine learning model (MLM), which is not developed earlier. MLM is integrate with traditional statistical model (TSM) for multiple linear regression (MLR) with the aims to identify significance of agronomical constrains (factors) for wheat production in Pakistan. The secondary data of crop cut experiments is collected from the Crop Reporting Service (CRS). Python’ key library (Scikit Learn) is used to analyze the experiment. The MLM is applied using 80% and 20% randomized partition. The MLM performed better than TSM for MLR. To find the better model, Cobb-Douglas regression is applied using the MLM and TSM for the same dataset. MLM is applied for the train dataset. Highest R² and lowest MSE and MAE found for Cobb-Douglas regression using MLM, comparing with TSM. The modified Cobb-Douglas regression using MLM found better fitted model. There is positive but insignificant relation exist for urea, water and adoption of new varieties trends. The harvesting and sowing period shows positive and significant relation, while DAP, other fertilizers, spray pest and soil type shows positive and highly significant relation against wheat productivity. This study can provide deep insights the productivity enhancement practices and can lead to layout the effective strategies to enhance wheat production with the aim to ensure the food availability.