https://yuanwanmydemo.mybluemix.net/
- In response to the COVID-19 pandemic, people from all kinds of science domain and industry domain work gother and try their best to figure out the problem faced by all over the world. Global research has been taken to apply recent advances in machine learning techniques to generate new insights in support of the ongoing fight against this infectious disease.Even as young software engineering graduates in SJSU, we are trying to do our best in this challenging time.
-
We want to build a ML model to predict the trend of the disease, which help to research on the future study of infectious deseases. And also, if accurate, this model can help people to prepare for COVID19, such as how long they will stay at home in different areas, how many people can be infectious in the future, which are can be dangerous and which are not.
-
We will estanlish a web site to offer the prediction of everywhere we can find enough data. People can select the area which they pay attention to and select the date they care about. For example, as the picture bellow:
-
This is a screenshot from the front of our project. The user can select the date in the future and it will demonstrate the prediction of the future development of COVID19.
-
As for prediction part,user can easily choose the future date,the level of quarantine to predict data
-
Also, we provide the current situation of COVID19
can get the newest information from our application.
- Data Set: We collect data from kaggle and github. The data source link is at the end of this file.
- Front End:react.js,boostrap,;diagram will be simpler and more intuitive to show data relation, as for chart showing,we choose echart.js to show our data.
- Back End:python,node.js,IBM cloud;we will collect plenty of data associated with COVID-19,then choose most suitable model, train our data in IBM Watson machine learning,then provide a API for predicting data.
- we fetch the latest daily data about covid-19 in every state and the nearest one month in USA from Intnet,then update them in our front page,so users
- IBM Cloud Machine Learning
- IBM Cloud Data Storage
- AWS
- ReactJS
- NodeJS
- Python
- https://www.politico.com/interactives/2020/coronavirus-testing-by-state-chart-of-new-cases/
- https://www.kaggle.com/sudalairajkumar/covid19-in-usa
- https://covid-19.direct/county/CA/Santa%20Clara
- Air quality historical data: https://aqicn.org/data-platform/register/
- https://www.kaggle.com/lanheken0/community-mobility-data-for-covid19
- https://www.kaggle.com/tarunkr/covid-19-case-study-analysis-viz-comparisons
- https://www.kaggle.com/sudalairajkumar/covid19-in-italy#covid19_italy_region.csv
There is a joke on a game Reddit that there are two players argue for their governments' reaction and solution correspond to Covid-19. And the story background is that the U.S just exceeded China and become the first country which have the most case of Covid-19.
- Player A: I am so proud to be the citizen of our country!
- Player B: XXXX XXXX(Some dirty adjective) Chinese.
- Player A: Dude, I am an American…
- Player B: …
We all knew that China was the only country which implemented quarantine rule and compulsorily required every citizens to wear mask at the virus exploding period. China had most Covid-19 cases in all over the world, however now China already controlled and weaken Covid-19's spreading speed in China area and set up plan to resume work and resume cities’ daily operation.
So, there are two questions may arise: Is wearing mask helpful for us to prevent getting the Covid-19 Virus? Does the quarantine rule works for control or weaken the Covid-19 positive cases' increase for a country?
I have traced back some previous news about the implementation of quarantine rule and the recommendation correspond to CDC/government suggested citizens to wear mask as necessary.
According to news' date, we could assume that the start date to implement the quarantine rule is Mar. 17th, and the date for CDC recommended people wearing facial masks is April 3rd. The dates around these two date (pre or post 2-3 days) also can be considered as the beginning of the implementation of rule & suggestion. Then we will check whether data give us clues from these two dates.
Below is the data about grocery & retailer mobility related to date. The higher mobility explained the more people go to grocery & retailer compared to last day. From the data, we can find that, the mobility increased dramatically from Mar.11th to Mar.17th, which also match the panic period in history-—people purchased over amounts of grocery to stay at home for long time. After Mar.17, we can find that the mobility begin to decrease, and the citizens did efficiently cooperate with government’s rule.
To make the data more readable, we will fix the fluctuation on the dataset. We will calculate the positive increase ratio per day. For example, Day 1, the increase number is 100, Day 2 is 50. Then the increase ratio should be -50%. etc…
According to the dataset, we assume that the covid-19 explode date for the US nation should be on Feb.28th. Before Feb.28th, the number of Covid-19 case still a few. Then we can skip till Feb 28th.
The trend of increase rate is not stable, it will be influenced by each day’s increase cases dramatically. It would be better for us to count the average to get a period’s increase rate. Then we can compare the data between the period before quarantine and after the quarantine. And the period before promoting mask VS after promoting mask.
And we can view that, after the government implements quarantine rule, the increase positive ratio got obvious decrease. From 40-50% down to 10-20%, we can view that the covid-19 positive case increase speed get slow down. More importantly, after April 3rd, the covid-19 positive case increase ration from 10-20% down to below 5%. It is very convincing that quarantine rule/promoting mask have positive effect for a country to fight with Covid-19.
The negative increase of the COVID-19 positive case does not mean the US nation already pass the turning point. Actually, the US nation still have not passed the turning point, though some state already passed the turning point.
I have implement a turning point algorithm. Let Just take Italy as the example.
Because the new increase case get influenced everyday, we can count the average increase cases as a period first. We can set the ratio as 3-day a period and 5-day a period. 3-day period could get more in precise date for turning point. However, if the data fluctuate in very huge range, we would recommend to use large scale range.
From the turning point algorithm, we know that, the turning point will be appear around the 30-32 row of the dataframe.
Which is between the dates from 3-25 to 3-27.
and we do find it the increase number vector get decrease after the turning point.
IF we implement this into the US dataset:
We will find it the turning point for the US nation not come yet, but at the period from April 6 to April 10 will have the most increase case per day. After that, later new COVID-19 cases per day have negative increase with very low rate.
Why quarantine rule and wearing mask work, why we still have lot of increase case per day? Previous work, such as the quarantine rule and wearing mask, we just control the increase rate for new COVID-19 cases. If we do not implement those works, right now, we may have not rather 30000 new cases per day, but 50000, 60000 new cases per day.
Why we still don’t have turning point yet?
Although for some state that already have turning point, the U.S is a very big country, that Covid-19 condition is different from states to states. Until all the states from U.S pass the turning point, then the whole nation will pass the turning point then.
In conclusion, we just control and reduce the accelerate rate for new cases per day, however there is still a distance for us to ultimately beat the virus for the whole nation.