Red line shows the true line for Cali 's deaths if we follow our current path to follow CDC. The true-labels relatively match predictions.
Shows the num deaths that will occur on our ability to wear masks. This would influence many people by showing them the death trend in Cali
Not apart of my video, but it contains the info about my implementation of CNN. This was my first time, so I was new on the terminology
I was watching the news and I found that many people did not want to wear masks and they believed that it did not "effect" the curve regarding the number of deaths regarding COVID-19. I realized that if more people had this philosophy then the pandemic would never come to an end. So I came up with the idea to show them how their undesirable decisions effect the number of casualties regarding the pandemic.
What it does
This architecture will use John Hopkins data to determine the future number of deaths regarding on people's ability to follow CDC guidelines on wearing a mask. Furthermore, the results of this code will show a graph that displays the future number of deaths for each state depending on each scenario on following CDC guidelines. This will show people in the US(a hotspot for coronavirus cases) how important and crucial it is to wear masks and follow CDC guidelines.
How I built it
I built this architecture using linear regression models and CNN layers for accuracy. I grouped all the data from Github into training and test sets and organized it with states as it was first plot within each county. I created an algorithm that would take the past 7 days and predict the next 14 days without using networks like RNN. I then further added weights to sort of smooth the linear regression curve. Furthermore, I added CNN layers to make the curve more smooth without the outliers. Then I added multiple CNN layers such as dense, convolutional, and maxpooling layers in order to make my mean squared error low.
Challenges I ran into
As I am a beginner at machine learning, it was a lot harder for me to adapt to the syntax of the linear regression model and the CNN network. Consequently, I ran into multiple challenges while making this project. For instance, I ran into multiple challenges while creating and applying my algorithm. My code would often get very complicated where I would have to go back to the code and traverse through it again which would take a long time. Moreover, I would have to constantly create multiple matrices in order to balance my shape and to add data into my architecture which would get more complicated as time went on which set me back a lot.
Accomplishments that I'm proud of
I am very proud of my architecture as I was able to properly apply my knowledge on deep learning and sklearn models into my code. Even thought this architecture took a lot of time and energy, I was very happy with my results as I was properly able to implement my CNN layers to make my root mean square error relatively low. Furthermore, I found ways to add more weights as to match the true labels with my predicted labels.
What I learned
I learned a lot while working through my project. For example, I learned to always keep in mind the shape of your matrices as you apply them and interchange them with other matrices. I would have simple matrices containing data of the cases and deaths per state be like (2 X 14), and later they would end up being(112, 14) which would often get frustrating as I would have to go through the code again. By knowing the shape of my matrices I was able to often track my mistakes as I would realize that I implemented my algorithm wrong.
What's next for "COVID-19: Predicts the future deaths using past data"
I was also thinking later to add RNN syntax into my code in order to further increase the accuracy. Furthermore, I want to apply this algorithm to bring awareness to other social problems in the world such as the starvation issue occurring in Yemen that not a lot of people are doing much about. I think that this algorithm can bring awareness to many people around the world as a means to encourage them to stand up to all parts or regions of the world as they are being severely impacted.