A CoCoSo-based decision support framework for electric vehicle evaluation using deep learning-derived policy sentiment evidence

Abstract

Electric vehicle adoption is shaped not only by technical and economic attributes but also by public perception of government incentive policies. Existing electric vehicle evaluation models commonly prioritize conventional criteria, while policy-related public sentiment is often treated only as descriptive context. This study proposes a hybrid decision support framework that integrates IndoBERT-based sentiment analysis with the Combined Compromise Solution (CoCoSo) method for electric vehicle evaluation in Indonesia. Public comments on electric vehicle incentives were classified into positive, neutral, and negative sentiment to construct a Policy Sentiment Evidence Index (PSEI). The index was incorporated into the CoCoSo decision matrix together with price, driving range, battery capacity, and charging time. The sentiment classifier achieved an accuracy of 0.6447 and a macro-F1 score of 0.5228, indicating moderate class-balanced performance under imbalanced sentiment distribution. The rescaled PSEI value of 0.3746 indicates relatively unfavorable public sentiment toward electric vehicle incentives. The CoCoSo results ranked Tesla Model 3 first, followed by Hyundai Ioniq 5, Nissan Leaf, and Wuling Air EV. Sensitivity analysis confirmed stable rankings across policy sentiment weighting scenarios, suggesting that sentiment evidence can enrich multi-criteria electric vehicle evaluation without destabilizing the decision outcome.