Downloads: 6 | Views: 95 | Weekly Hits: ⮙2 | Monthly Hits: ⮙5
Analysis Study Research Paper | Computers in Biology and Medicine | India | Volume 12 Issue 8, August 2023
Utilizing Machine Learning Algorithms for Analysing and Forecasting COVID-19 Pandemic Data
Dr. R Amutha | Dr. S. Karthik [2]
Abstract: India's inaugural COVID- 19 case was documented on January 30th, 2020, and the incidence of reported cases surged significantly from March 2020 onwards. This research paper undertakes an extensive analysis of COVID-19 data, commencing at a global scale and subsequently narrowing down the focus to India's context. The dataset is sourced from multiple reliable government websites, ensuring data authenticity. The urgency lies in accurately projecting the point of peak cases and their subsequent decline. This information holds immense value for public welfare professionals in strategizing preventive measures while balancing economic considerations. Python and Data Visualization techniques are employed to depict variables such as gender, geographical distribution, and age demographics. Time Series Forecasting techniques, encompassing Machine Learning models like Linear Regression, Support Vector Regression, Polynomial Regression, and a Deep Learning Forecasting Model-LSTM (Long short-term memory), are harnessed to scrutinize potential surges in cases both in the near and distant future. A comparative evaluation is executed to discern the model that aligns most fittingly with the data. This research paper constructs predictive models geared towards anticipating positive case counts with heightened accuracy. By leveraging Regression-based, Decision tree-based, and Random forest-based models established on China's dataset and cross-validating them with India's sample, the efficacy of the models is evident. Their ability to project future positive case numbers with minimal error is established. The resultant machine learning model operates in real-time, proficiently predicting positive case counts. Significantly, the paper advances critical measures and recommendations in the context of lockdown's impact. To enhance data comparability and mitigate data extremities or outliers, certain feature engineering techniques transform the data into logarithmic scales. The model's predictive capability is twofold: short-term intervals are projected, and the model's adaptability for long-term forecasting is acknowledged and can be fine-tuned.
Keywords: COVID 19, Data Analysis, Forecasting, Machine learning, Deep Learning, Sigmoid Curve, LSTM Model
Edition: Volume 12 Issue 8, August 2023,
Pages: 2095 - 2101