Time-Series-Forecasting-For-Covid-19

Forecasting Covid-19 Cases in India

Aim

To build a model that can predict the values of upcomming Covid-19 cases provided the timeseries of already confirmed datewise cases in India. The approach used in this project can be exteded to statewise data to get statewise predictions.

Methodology

1. Gathering data from API ( 'https://api.covid19india.org/csv/latest/state_wise_daily.csv' ) and preprocessing.
2. Analysing if current time series is stationary or not using Augmented Dickey-Fuller unit root test.
3. Converting the series into stationary using differencing method.
4. Analysing ACF and PACF and tuinning values of p,d,q.
5. Using our model to predict the values of the current timeseries and comparing it with orignal data.
6. Forecasting and plotting values for future.

Results

Comparing predicted value to orignal values

Forecasting for upcomming week

References

1. https://www.kaggle.com/sumi25/understand-arima-and-tune-p-d-q
2. https://machinelearningmastery.com/time-series-forecasting-methods-in-python-cheat-sheet/
3. Api for gathering data : https://api.covid19india.org/