This study is to solve the problems of an overly-broad scale of medical indicators, lack of retrospective research samples, insufficient depth of data mining, and low disease prediction accuracy. In this paper, we propose an intelligent screening algorithm that combines a genetic algorithm, cellular automata, and rough set theory. This algorithm can achieve high accuracy in predicting patient outcomes with a small number of indicators. And we compare it with the traditional genetic algorithm. We built the prediction model with 64 indicators based on the logistic regression (AUC 0.8628), support vector machine (AUC 0.5319), Naïve Bayes (AUC 0.7102), and AdaBoost algorithms (AUC 0.9095). Using the cellular genetic algorithm for attribute screening not only effectively reduces the number of indicators but also achieve almost the same accuracy of prediction with 8 indicators based on the logistic regression (AUC 0.8782), support vector machine (AUC 0.8525), Naïve Bayes (AUC 0.8408), and AdaBoost algorithms (AUC 0.8770). Compared with the traditional scoring system, the predictive model established in this paper can more accurately predict rebleeding accidents based on physiological test indicators and continuous patient indicators.

中文翻译：

---

粗糙集和细胞遗传融合算法在急性重大疾病预测中的应用

本研究旨在解决医学指标规模过大，缺乏回顾性研究样本，数据挖掘深度不足以及疾病预测准确性低的问题。在本文中，我们提出了一种智能筛选算法，该算法结合了遗传算法，元胞自动机和粗糙集理论。该算法可以通过少量指标实现高精度的患者预后预测。并将其与传统的遗传算法进行比较。我们基于逻辑回归（AUC 0.8628），支持向量机（AUC 0.5319），朴素贝叶斯（AUC 0.7102）和AdaBoost算法（AUC 0.9095）建立了具有64个指标的预测模型。使用细胞遗传算法进行属性筛选不仅可以有效地减少指标数量，而且还可以基于逻辑回归（AUC 0.8782），支持向量机（AUC 0.8525），朴素贝叶斯（AUC）使用8个指标实现几乎相同的预测准确性0.8408）和AdaBoost算法（AUC 0.8770）。与传统的评分系统相比，本文建立的预测模型可以根据生理测试指标和连续患者指标更准确地预测再出血事故。