توصيفگر ها :
يادگيري ماشين , جنگل تصادفي , انتخاب ويژگي بازگشتي , كوويد-19 , تحليل داده‌هاي پزشكي
چكيده فارسي :
بيماري كوويد-19، ناشي از ويروس (2SARS-CoV) تأثيرات گسترده‌اي بر سلامت عمومي در سطح جهاني داشته است. از اين رو به تحليل جامع داده‌هاي باليني و آزمايشگاهي بيماران مبتلا به كوويد-19 با هدف درك عميق از ويژگي‌هاي جمعيت شناسي، باليني و آزمايشگاهي اين بيماري از اهميت فراواني برخوردار است.
در اين تحقيق داده هاي باليني بيماران مبتلا به كويد-19 مورد بررسي قرار گرفته است و مشخصات باليني، يافته هاي آزمايشگاهي، درمان ها و نتايج مورد تحليل قرار گرفته اند.
هدف تحقيق پيش بيني دقيق از تشخيص باليني بيماري بوده است. نتايج اين تحقيق نشان مي‌دهد كه شاخص‌هاي آزمايشگاهي نظير CRP و D-Dimer در پيش‌بيني نتايج باليني و مديريت بيماري مؤثرند و تفاوت‌هاي معناداري در ويژگي‌هاي باليني و آزمايشگاهي بين بيماران ترخيص شده و فوت شده وجود دارد. اين يافته‌ها مي‌تواند به بهبود استراتژي‌هاي درماني و پيشگيري از بيماري كوويد-19 و توسعه پژوهش‌هاي آينده در اين حوزه كمك كند.
اين تحقيق با استفاده از روش‌هاي يادگيري ماشين انجام شده است. از الگوريتم‌هاي مختلف يادگيري ماشين شامل جنگل تصادفي و تكنيك انتخاب ويژگي بازگشتي (RFE) براي تحليل داده‌ها استفاده شد. اين روش‌ها به شناسايي ويژگي‌هاي كليدي و مدل‌سازي روابط پيچيده بين متغيرها كمك كردند تا پيش‌بيني دقيق‌تري از نتايج باليني بيماران كوويد-19 ارائه شود.
در تحقيق ديگر پيش بيني موارد تائيد شده، درمان شده و فوت شده از بيماري كويد-19 به عنوان يك وظيفه طبقه بندي چندكلاسه مورد بررسي قرار گرفته است و از طريق روش هاي متداول يادگيري بانظارت مانند ماشين بردار پشتيبان، درخت تصميم، شبكه عصبي و جنگل تصادفي نتايج بررسي و مقايسه شده است.
چكيده انگليسي :
This thesis examines and analyzes the clinical and laboratory data of COVID-19 patients using advanced machine learning methods. Given the global spread of SARS-CoV-2, which led to the COVID-19 pandemic, and its widespread impact on healthcare, economic, and social systems worldwide, identifying key predictors of clinical outcomes is of paramount importance. The primary goal of this research was to develop a machine learning-based model for more accurate prediction of patient outcomes and to improve their treatment management as well as performing timely preventive actions. To achieve this, the random forest algorithm (RF) and the recursive feature elimination (RFE) technique were employed to design and implement a comprehensive and efficient model for predicting clinical outcomes in COVID-19 patients.
The research involved several essential steps. First, clinical and laboratory data were collected from 126 COVID-19 patients treated at Wuhan's Fourth Hospital. These data included clinical variables such as age, gender fever, and laboratory results like CRP, D-dimer, and other relevant markers. Using the random forest algorithm, a predictive model was developed that could accurately forecast the clinical outcomes of these patients. To further optimize the model’s performance and improve prediction accuracy, the RFE technique was employed to identify key variables and reduce the number of features. This process led to a reduction in model complexity and enhanced its efficiency. The results of the study revealed that certain clinical markers, such as LDH and Myo levels, were identified as critical predictors of patient outcomes and played a significant role in determining treatment outcomes for COVID-19 patients.
This study also compared the performance of the random forest algorithm with other machine learning models. To this end, a model was developed to predict confirmed, recovered, and deceased COVID-19 cases. The data for this section were collected from various regions of India, and a more accurate prediction model was generated using the random forest algorithm. This model was compared with other machine learning models such as support vector machines (SVM), decision trees, and neural networks. The results showed that random forest outperformed these other models in predicting COVID-19 cases. These findings demonstrate the power and efficiency of machine learning algorithms in forecasting disease trends and in enhancing public health planning and management for combating pandemics.
The findings of this research underscore the significant value of using machine learning methods, especially the random forest algorithm, in predicting the clinical outcomes of COVID-19 patients. These methods can not only increase prediction accuracy but also help identify key variables that can lead to the development of more effective treatment strategies. In this study, the random forest algorithm successfully provided a comprehensive and accurate model that offered higher prediction precision compared to other available methods. This enables healthcare professionals to make better clinical decisions and manage patients more effectively. In fact, one of the primary contributions of this research is providing a tool for early prediction of patient outcomes and improving the management of complex diseases like COVID-19.
