International Journal of Computer Applications (0975 – 8887) Volume 184– No.21, July 2022 60 A Real-time Model to Forecast the Outbreak of Covid-19 using LSTM Omkar Bhoite Modern Education Society’s College of Engineering Pune, India Sohail Ahmad Modern Education Society’s College of Engineering Pune, India Saurabh Wagh Modern Education Society’s College of Engineering Pune, India Ketan Gaikwad Modern Education Society’s College of Engineering Pune, India Shalaka Deore Modern Education Society’s College of Engineering Pune, India Shubhangi Ingle Modern Education Society’s College of Engineering Pune, India ABSTRACT Deep Learning based forecasting models have been in use for a long time and they have proven their significance in problems including time series forecasting and improve the accuracy and efficiency of the results for given problem. These models have long been utilization domains that required that identification and fiction of main factors of information. Based on review of work done in the field of forecasting, this study demonstrates the potential of LSTM algorithm to forecast the rise and fall in number of active cases and deaths of Covid 19 patients using real time input provided by John Hopkins University which is available on GitHub and updated on a daily basis. In short, a real time Covid 19 outbreak forecasting model implemented using long short term memory networks algorithm. The use of LSTM is suggested to improve the efficiency and accuracy of the presently available models and make predictions of 2 parameters including the number of active cases and the number of deaths for the upcoming 10 days. The goal was to analyze the algorithm by comparing the results of prediction and actual reports for a period of 60 days and forecast the number of newly confirmed and death cases of the disease for upcoming 10 days. To develop an algorithm faster than existing systems and use the most recently available data for a higher range of input and calculate the latest trend. Keywords LSTM, Optimizer, Activation function, MinMaxScaler, Dense Layer, Environment Variables 1. INTRODUCTION This study reviews some of the state of art supervised machine learning regression models including least absolute linkage and selection operator (LASSO), supporter machine and exponential smoothening. These models have been trend using co-ed 19 station statistics data set provided by the John Hopkins University. The data has been preprocessed and divided into two subsets: training set and testing set. As of now, the dataset contains more than 850 entries out of which, the last 60 entries are reserved for testing the data. In the current human crisis, our main attempt is to develop a forecasting system for Covid 19 which will help in implementing several policies for the government and curtail its spread. The timely decisions taken by the government helped in controlling the speed of Covid 19 to a large extent. Despite these decisions the pandemic continued to spread. Today, it seems we have restricted the spread but this system will be helpful to predict upcoming waves and take desired actions. To improve your experience and increase the efficiency of the system we have developed a front end considering various UX design experience aspects. The design is simple but effective. It gives a general idea of the virus, the authors, provides latest covid-19 stats, predicts outbreak of 19 for the next 10 days and shows a graph of collected and predicted data for both the parameters. 2. MODULE IDENTIFICATION Algorithm: LSTM algorithm Input: Automatically downloaded CSV file from GitHub Output: Prediction results and comparison graphs 3. LITERATURE SURVEY In this paper [1], during this paper the algorithm used is simple regression. Prediction and outbreak are taken into account to be a regression problem and it's implemented using two regression models namely linear and polynomial regression. The Covid-19 India data set is used for implementation purposes it predicts the number of confirmed new cases, deaths and recoveries based on data available from 12th March 2:30 1st October 2020. for the forecasting purpose and analysis statistic forecasting approach is employed. The output of this implementation suggests that polynomial regression shows better results than simple regression. The forecasting was done using Tableau and also the results are satisfactory, but the results may be more accurate upon use of a much bigger dataset considered for an extended duration. In the paper [2], Supervised Machine Learning Model algorithm is employed for the prediction in this paper. Machine learning (ML) based forecasting mechanisms have proved their significance in forecasting outcomes to improve the decision making on the future course of actions. Such models have been used for a protracted time in many applications which needed the identification and prediction of any sort of adverse factors or analyzing threats. Several prediction methods are being used to handle forecasting problems. This study demonstrates the potential of ML models to forecast the number of upcoming