Done by Prannaya Gupta
In this project, we analyze the Singapore Covid-19 data using multiple Time Series Models. The prediction analyses data from online and runs different models to find an optimum prediction for the Covid-19 Slope. We predicted the COVID-19 outbreak trajectory for Singapore in the next 6 months (from September 2020 this is very outdated). We retrieved data from OWID and used ARIMA, a modified version of Double Exponential Smoothing and the Prophet API to test and evaluate the data, and made predictions. This project was done as part of the Singapore Problem X (PX) Challenge. You can look into it here.
As the name suggests, this section deals with importing important libraries, retrieving the data from the OWID Github .csv file. Here, we also account for the time lag in the recording of data by OWID.
Here, we test the data in order to find out many different possible curves to use for the prediction. Click here for info regarding why this series of tests is conducted and what the significance is.
We use the ARIMA method to predict the best possible with the least MAPE score. Once again, click here for info regarding the significance of this test.
This is a modified algorithm for Time Series Analysis. This algorithm was made by me and does not follow any normal method. I use my data from the double exponential smoothing test and used the very accurate curve to predict to a further extent. My algorithm uses trends and levels from the previous 10 days.
I use Facebook's Prophet API to generate a model. This was mainly inspired by info from here.
Here, I used all previous data to average out and make an overally data prediction. The cases individually until October 1st are all laid out and there is also an accompanying graph.
Open the document in colab since there are a lot of unnecessary datasets that cannot be minimised in Github.