<p>Prediction for enterprise credit risk in the era of Industry 4.0 is important for early credit crisis warning. Existing evaluation methods for small and medium-sized enterprises (SMEs) credit levels simply using financial information have two major limitations. Firstly, it relies heavily on financial reports to ignore the soft information, including the influence of digital supply chain. (DSC). Secondly, it demands higher efficiency in feature selection with the increase in the dimension of credit data. Therefore, we introduce a novel ensemble model for SMEs credit risk prediction combined with the improved Artificial Bee Colony (ABC) algorithm for feature selection (IABCFS). Firstly, we build a new evaluation system for SMEs credit risk with consideration of the information value of digital supply chain, and then use TF-IDF algorithm to measure the indexes. Secondly, we improve the standard ABC algorithm for feature selection by a new search strategy, and use Symmetric Uncertainty (SU) to construct the fitness function for the higher efficiency. Finally, we introduce the SVM-Adaboost ensemble model to generate the prediction result, and use SMOTE algorithm with Tomek Link Removal to handle the problem of class imbalance. The results of feature selection show that the information of DSC in SMEs credit risk prediction deserve more attention, and our proposed model also achieves competitive results both on a Chinese SMEs dataset and three public datasets.</p>