Coronavirus (COVID-19) Prediction

This project aims to predict the numbers that are published in each day regarding the amount of Coronavirus (COVID-19) cases and deaths.

Table of contents

• Requirements
• How to use
• How does it work
• Contact info
• Adding models
• Output examples
  • Terminal output
  • Cases in each day graph
  • Deaths in each day graph
• Predictions

Requirements

1. A machine with Python 3 installed.
2. The following packages are needed to be installed for this project to run:
   • numpy
   • matplotlib
   • scikit-learn
   • BeautifulSoup

How to use

(Using Poetry)

1. Install Poetry
2. Run poetry install
3. Run poetry run python main.py

(Not using Poetry)

1. Install dependencies with pip install -r requirements.txt
2. Run python main.py

How does it work

The main.py file uses the DataGrabber class (source included) to fetch the required data from https://www.worldometers.info.
The main.py file then trains 2 polynomial models using the fetched data and scikit-learn's LinearRegression - the cases in each day model and then the deaths in each day model.
Afterwards, the file displays the models' predictions for the next day, the functions that depict the trained models and displays a graph for each model.

Contact info

You may contact me via Linkedin: https://www.linkedin.com/in/eladcn/.

Adding models

It is possible to add more models (e.g. models for specific countries) to the project by taking the following steps:

1. Create a new DataGrabber class (you can name it however you want, for example: USADeathsDataGrabber), it doesn't have to inherit from class DataGrabber.
2. Implement the following method in the new class you created: def get_dataset_file_name(self, dataset_date) This method should return the dataset file name for a specific given date. For example, it may return: "USA_" + dataset_date + ".csv". You can find a good example for this in the CasesDataGrabber class.
3. (Optional) If you can fetch the data from an external source (instead of managing the dataset manually) and you would like to implement this feature, implement the following method in the new class you created: def grab_data(self) This method should fetch the data from a data source and store it into a file (which is the same format as mentioned in section 2).
4. Add the following configuration to the models array inside the config.json file:

        {
            "enabled": true, // Whether the model is enabled and the program should handle it or not.
            "model_name": "USA Deaths", // This is only used for display reasons and will not affect any logic.
            "model": {
                "type": "regression", // The model type, could be either "regression" or "neural_net".
                "polynomial_degree": 6 // A hyper parameter for regression models. See the config file for a neural network example.
            },
            "datagrabber_class": "USADeathsDataGrabber", // The class we would like to use.
            "grab_data_from_server": true, // Set this to false if you would like to manage the dataset manually (i.e not use the grab_data(self) method).
            "offline_dataset_date": "2020-04-10", // If the dataset is managed manually - specify the date of the offline dataset file.
            "days_to_predict": 10 // How many days ahead you would like to predict.
        }

Please note that the data should be formatted in a CSV file with the following structure:

0,value_for_day_0
1,value_for_day_1
2,value_for_day_2
.
.
.
n,value_for_day_n

Output examples

Terminal output

Cases in each day graph

Deaths in each day graph

Predictions

Date	Cases	Deaths	Predicted Cases	Predicted Deaths	Notes
13.05.2020	4,425,655	297,765	4,400,686	305,157
12.05.2020	4,337,602	292,451	4,315,787	298,811
11.05.2020	4,252,290	287,137	4,218,443	294,230
10.05.2020	4,178,154	283,734	4,132,764	290,529
09.05.2020	4,098,288	280,224	4,045,221	285,381
08.05.2020	4,009,291	275,976	3,961,661	279,957
07.05.2020	3,913,644	270,426	3,875,276	274,533
06.05.2020	3,817,382	264,837	3,790,035	266,889
05.05.2020	3,724,518	258,027	3,707,471	263,849
04.05.2020	3,643,271	252,241	3,616,918	256,777
02.05.2020	3,481,429	244,665	3,463,870	246,550
01.05.2020	3,398,473	239,448	3,365,306	241,043
30.04.2020	3,304,220	233,830	3,290,765	234,227
29.04.2020	3,218,184	228,030	3,205,781	224,419
28.04.2020	3,136,508	217,813	3,129,694	219,813
21.04.2020	2,556,720	177,675	2,553,102	174,137
20.04.2020	2,480,503	170,397	2,476,492	169,819
19.04.2020	2,406,575	165,031	2,396,306	163,820
18.04.2020	2,330,766	160,047	2,313,171	152,571
17.04.2020	2,248,863	154,145	2,247,654	148,412	Both models are now using a neural network architecture.
16.04.2020	2,181,334	145,471	2,153,255	138,052
15.04.2020	2,086,332	138,475	2,068,918	131,251	The cases were predicted using a neural network.
14.04.2020	2,001,681	130,379	1,990,227	125,759
13.04.2020	1,923,937	119,618	1,927,047	121,115
12.04.2020	1,852,365	114,196	1,857,691	115,824
11.04.2020	1,779,842	108,779	1,779,990	109,470	Changed the cases model's polynomial degree.
10.04.2020	1,698,881	102,687	1,662,881	102,308
09.04.2020	1,603,694	95,693	1,585,772	95,185
08.04.2020	1,518,023	88,457	1,508,529	88,618
07.04.2020	1,430,981	82,036	1,436,648	82,061
06.04.2020	1,346,036	74,654	1,369,235	77,013
05.04.2020	1,272,901	69,427	1,309,632	72,120	Changed the cases model's polynomial degree.
04.04.2020	1,201,483	64,691	1,199,951	65,524
03.04.2020	1,116,662	59,162	1,101,907	58,909	Restored the deaths model's polynomial degree.
02.04.2020	1,015,065	53,167	1,023,706	52,199	Changed the models' poloynomials degrees.
01.04.2020	935,197	47,192	965,240	47,590
31.03.2020	858,355	42,309	885,894	42,849
30.03.2020	784,794	37,788	810,279	38,570
29.03.2020	722,359	33,966	734,545	34,636
28.03.2020	663,124	30,862	657,417	30,684
27.03.2020	596,366	27,344	585,678	27,119
26.03.2020	531,865	24,073	521,527	23,997
25.03.2020	471,035	21,284	466,864	21,207
24.03.2020	422,599	18,894	417,312	18,651
23.03.2020	378,860	16,514	372,656	16,494
22.03.2020	337,469	14,647	334,355	14,544
21.03.2020	305,036	13,013	298,845	12,760
20.03.2020	275,598	11,387	265,940	11,271
19.03.2020	244,933	10,031	239,006	10,027
18.03.2020	218,822	8,951	217,680	8,934
17.03.2020	198,234	7,978	200,305	8,009
16.03.2020	182,473	7,160	184,930	7,223
15.03.2020	169,577	6,519	170,336	6,504
14.03.2020	156,622	5,833	157,620	5,966
13.03.2020	145,483	5,429	146,450	5,434
12.03.2020	134,577	4,982	137,917	4,984
27.02.2020	83,113	2,858	-	2,867
26.02.2020	81,828	2,801	-	2,839
25.02.2020	80,828	2,763	-	2,775
24.02.2020	80,087	2,699	-	2,679	The cases model needs to be changed.
23.02.2020	79,205	2,618	79,611	2,548
22.02.2020	78,651	2,460	80,423	2,458	Changed the cases model polynomial degree.
21.02.2020	77,673	2,360	75,162	2,355
20.02.2020	76,667	2,247	76,109	2,248	The data for previous days was changed in this day.
19.02.2020	75,700	2,126	77,427	2,138
18.02.2020	75,184	2,009	77,842	2,030	Changed the polynomials degrees.
17.02.2020	73,332	1,873	79,231	1,956
16.02.2020	71,329	1,775	76,943	1,824
15.02.2020	69,197	1,669	73,331	1,668
14.02.2020	67,100	1,526	67,496	1,516
13.02.2020	64,438	1,383	58,692	1,376
12.02.2020	45,134	1,261	48,123	1,233