Predicting the outbreak of disease is essential when managing shrimp farms. Acute hepatopancreatic necrosis disease (AHPND) caused by Vibrio parahaemolyticus is a serious disease in shrimp. It is essential that shrimp farmers on the east coast of the Mekong Delta detect the disease as early as possible, because the mortality rate can reach 100%. Here, we used machine learning to predict AHPND development based on data collected since 2010 from shrimp farms in Tra Vinh, Ben Tre, Bac Lieu, and Ca Mau provinces. We initially hypothesized that the dependent variable, AHPND, was affected by 31 independent variables, but ultimately used 15 key variables to train the models. Logistic regression, artificial neural network, decision tree, and K-nearest neighbor analyses were performed, and the accuracy of the predictions was evaluated using hold-out and cross-validation tests. Logistic regression, as the most stable algorithm, was thus used to predict AHPND outbreaks in shrimp farms.

中文翻译：

---

使用机器学习预测越南湄公河三角洲东海岸养殖虾的急性肝胰腺坏死病 (AHPND)

在管理养虾场时，预测疾病的爆发至关重要。由副溶血性弧菌引起的急性肝胰腺坏死病（AHPND）是对虾的一种严重疾病。湄公河三角洲东海岸的虾农尽早发现疾病至关重要，因为死亡率可达100%。在这里，我们使用机器学习根据自 2010 年以来从茶荣省、槟知省、薄辽省和金茂省养虾场收集的数据来预测 AHPND 的发展。我们最初假设因变量 AHPND 受到 31 个自变量的影响，但最终使用 15 个关键变量来训练模型。进行了逻辑回归、人工神经网络、决策树和 K-近邻分析，并且使用保持和交叉验证测试来评估预测的准确性。因此，逻辑回归作为最稳定的算法被用于预测养虾场的 AHPND 爆发。