The financial condition of an enterprise is a critical factor in determining its viability and long-term prospects, and it also directly affects the interests of stakeholders such as investors and creditors. So extra attention is needed to evaluate and forecast the financial risk of an enterprise. This paper proposes a novel enterprise financial risk evaluation and prediction model integrating Synthetic Minority Over-sampling and Edited Nearest Neighbors (SMOTE-ENN) and Natural Gradient Boosting (NGBoost). The CRITIC-TOPSIS method is first employed to conduct a comprehensive evaluation of the financial risk of the listed companies based on their annual financial statements. Subsequently, the obtained evaluation results are subjected to a clustering analysis to categorize the companies into different levels of financial risk. To address the issue of imbalanced samples, the study introduces the SMOTE-ENN combined sampling method. This technique is applied to mitigate the imbalance in the number of samples across different categories. Lastly, NGBoost is utilized to develop a prediction model for corporate financial risk based on the clustered data. The analysis reveals that companies with low levels of financial risk typically exhibit a balanced financial profile, robust operational performance, and substantial growth potential. Conversely, companies with high levels of financial risk are characterized by weak profitability and debt-paying ability. The experimental findings show that the model has a 97.18% accuracy rate, thereby suggesting that the model is both reasonable and practicable.

We propose a DeiT (Data-Efficient Image Transformer) feature encoder-based algorithm for identifying disease types and generating relevant descriptions of diseased crops. It solves the scarcity problem of the image description algorithm applied in agriculture. We divided the original image into a sequence of image patches to fit the input form of the DeiT encoder, which was distilled by RegNet. Then, we used the Transformer decoder to generate descriptions. Compared to “CNN + LSTM” models, our proposed model is entirely convolution-free and has high training efficiency. On the Rice2k dataset created by us, the model achieved a 47.3 BLEU-4 score, 65.0 ROUGE_L score, and 177.1 CIDEr score. The extensive experiments demonstrate the effectiveness and the strong robustness of our model. It can be better applied to automatically generate descriptions of similar crop disease characteristics.