Ngabo Desire, Dong Wang, Ibeke Ebuka, Iwendi Celestine, Masabo Emmanuel
College of Computer Science and Electronic Engineering, Hunan University, Changsha 410000, China.
African Center of Excellence in the Internet of Things, University of Rwanda, Kigali 3900, Rwanda.
Math Biosci Eng. 2021 Sep 27;18(6):8444-8461. doi: 10.3934/mbe.2021418.
With the recent advancement in analytic techniques and the increasing generation of healthcare data, artificial intelligence (AI) is reinventing the healthcare system for tackling pandemics securely in smart cities. AI tools continue register numerous successes in major disease areas such as cancer, neurology and now in new coronavirus SARS-CoV-2 (COVID-19) detection. COVID-19 patients often experience several symptoms which include breathlessness, fever, cough, nausea, sore throat, blocked nose, runny nose, headache, muscle aches, and joint pains. This paper proposes an artificial intelligence (AI) algorithm that predicts the rate of likely survivals of COVID-19 suspected patients based on good immune system, exercises and age quantiles securely. Four algorithms (Naïve Bayes, Logistic Regression, Decision Tree and k-Nearest Neighbours (kNN)) were compared. We performed True Positive (TP) rate and False Positive (FP) rate analysis on both positive and negative covid patients data. The experimental results show that kNN, and Decision Tree both obtained a score of 99.30% while Naïve Bayes and Logistic Regression obtained 91.70% and 99.20%, respectively on TP rate for negative patients. For positive covid patients, Naïve Bayes outperformed other models with a score of 10.90%. On the other hand, Naïve Bayes obtained a score of 89.10% for FP rate for negative patients while Logistic Regression, kNN, and Decision Tree obtained scores of 93.90%, 93.90%, and 94.50%, respectively.
随着分析技术的最新进展以及医疗保健数据的不断增加，人工智能（AI）正在重塑医疗保健系统，以便在智慧城市中安全地应对大流行病。人工智能工具在癌症、神经学等主要疾病领域以及现在的新型冠状病毒SARS-CoV-2（COVID-19）检测方面继续取得众多成功。COVID-19患者通常会出现多种症状，包括呼吸急促、发烧、咳嗽、恶心、喉咙痛、鼻塞、流鼻涕、头痛、肌肉疼痛和关节疼痛。本文提出了一种人工智能（AI）算法，该算法基于良好的免疫系统、运动和年龄分位数安全地预测COVID-19疑似患者的可能存活率。比较了四种算法（朴素贝叶斯、逻辑回归、决策树和k近邻（kNN））。我们对阳性和阴性COVID患者数据进行了真阳性（TP）率和假阳性（FP）率分析。实验结果表明，kNN和决策树在阴性患者的TP率上均获得了99.30%的分数，而朴素贝叶斯和逻辑回归分别获得了91.70%和99.20%。对于阳性COVID患者，朴素贝叶斯以10.90%的分数优于其他模型。另一方面，朴素贝叶斯在阴性患者的FP率上获得了89.10%的分数，而逻辑回归、kNN和决策树分别获得了93.90%、93.90%和94.50%的分数。
