predicting the number of COVID cases, Deaths & number of Cured cases using Indian covid count dataset from kaggle.
This is a time series analysis problem
Worked on various types of visualization to understand how many cases/deaths/cured cases are being reported monthly, weekly and daily basis.
converted the dataset into time series using below method:
First divided the whole dataset into train and test data and converted them to numpy arrays.
Then for every 'n' number of samples in train data do
take first n-1 data points as X, and nth sample as y
For test data, take first n data set x_test, and remaining all as y_test
Used LSTM architecture to predict the number of cases
Then used statistical models for decomposition of Cured count into 3 time series components: Trend, Seasonality and Residual.
Used Augmented Dickey Fuller statistical test to find out whether the given Time Series is stationary (fixed mean and variance over time) or not. Here, if p-value is less than significance level (0.05), the we can reject the null hypothesis (the time series is not stationary)
Used statistical models like Holt, SimpleExponentialSmoothing to predict the future cases.
Use Case-2: Covid Presence Detection - Based on symptoms given as features
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent