Optimizing the Classification Assistance through Supply Chain Management for Telematics SMEs in Indonesia using Deep Learning Approach

Abstract

This study aims to optimize the classification process of providing assistance to Indonesian Telematics Small and Medium Enterprises (SMEs) using a deep learning approach. The data used is the 2016 Economic Census data. The research was conducted comprehensively through the process of comparing performance through several approaches. Deep learning performance shows an optimal accuracy rate of 99.03%, higher than other approaches of the Adaboost and Adaboos-Bagging Ensemble (92.0%), LVQ (93.11%) and Backpropagation (89.1%). The deep learning approach still has shortcomings in terms of tracing attributes that affect the provision of assistance. Unlike the case with an ensemble approach that is able to display priority attributes, and these results are also validated with relevant research results. Research development opportunities can be done through the integration of EXPLAIN and IME models in the deep learning model, making it easier for stakeholders to prioritize attributes that affect the delivery of telematics SMEs. This is expected to encourage the improvement of SMEs competitiveness in facing the challenges of the Industrial Revolution 4.0.