Efficiency Management of Women Poultry Farmers Using Hybrid DEA and Machine Learning Approach: A Case of SHG-based Production in Sub-Himalayan North Bengal

Abstract

Data envelopment analysis (DEA) offers a linear programming approach to evaluate the efficiency in diverse fields of production and service sectors with wide utilization for effective performance measurement operations. DEA has found its useful applications in agriculture to examine optimal resource use for sustainable consumption. The popularly used two-step process where DEA is employed along with a regression model to explain the impact of exogenous factors on efficiency has been employed in past studies. This article aims to combine the conventional DEA approach with machine learning (ML) models for establishing a novel alternative method for performance measurement as well as the prediction of key exogenous factors affecting the efficiency of the women self-help groups (SHGs) led poultry farmers in sub-Himalayan North Bengal surrounding the Siliguri region of Darjeeling district. For this purpose, in the first step, DEA was employed to measure the efficiency of 80 women poultry farmers belonging to 20 SHGs and in the second step, the state-of-the-art random forest (RF) technique has been employed to predict the most important efficiency influencing variables. The results suggested inefficiencies among the SHG women with wide variation between the efficient and inefficient units. The use of the RF model predicted important factors such as the role of non-governmental organizations, educational level, financial inclusion, landholding and poultry rearing experience in years to impact the efficiency of these women farmers. As a result, the hybrid DEA-ML approach is useful to tackle ill adversities in poultry production that may help the women SHGs to develop agriculture-based income.