CharmingApe
8 minute read
CharmingApe is a web service which provides Monkeypox-related information targeting Autralian male workforce who may invovle frequent skin-to-skin contact. The main objective of the project is to spread awareness and provide a prediction of monkeypox cases on daily basis based on our self-developed Machine-learning Model.
Type: Healthcare Sector
Problem Statement
Recently there has been a rise of monkeypox cases in Australia, whereby Victoria and New South Wales to this day make up 43.3% and 44.4% of cases in Australia collectively. Globally, the data reflects males to make up 98.87% of cases with more than 50% of these cases being within the age range of 20-69. However, the top 4 infected countries show that the majority of their cases are inferred to lie within ages of 25-34. Monkeypox has been stated to have a higher rate of infection via extensive skin to skin contact, with a lower rate of infection via fluid. With countless people working in the service industry requiring extensive periods of consistent skin to skin contact, increases their chances of catching the virus. How might we raise awareness to male Australians in this industry who are constantly exposed to extensive skin to skin contact about monkeypox and prevent them from getting the disease?
System Architecture
Serverless & Microservice
Backend was designed based on the microservice architecture following a serverless cloud-native development model. Function as a Service (FaaS) provides a platform ,which allows rapid build, run, test and deployment of an application. The characteristic of ‘separation of concern’ allows future update and maintenance to be performed relatively easily without having to interrupt any other functioning features.
Event-driven & FaaS
FaaS has been divided into two roles, with one responsible for retrieving data from the database and the other role responsible for updating the data in the database. Regarding the event-driven characteristics of FaaS, two types of trigger have been used. API calls, as the first trigger, is essentially a request from the frontend, which will be served as an event invoking the backend query functions for retrieving data from the database. Scheduled event rules, as the second trigger, was achieved using AWS CloudWatch to create tailored timed events, which will serve as the trigger of backend database update functions to update the requested table.
Machine-Learning Model
Size limit of FaaS was a big obstacle for the automation due to the fact that our team lacked of experience of working with self-developed ML model. Without the assistance from AWS, the most viable solution at that time was to utilise both the database and CRON-job scheduling (Linux native scheduling program) to achieve daily prediction of Monkeypox cases in AU. With database as the middleware for frontend-tier to request and CRON-job as the scheduling tool to trigger the model runs and inserts the results into the database daily.
Deployment
To optimise both the cost and performance for small-scale application, the serverless architecture was deemed as the right fit for the backend-tier. AWS Lambda automatically handle scaling from a single request to hundreds of thousands per second. The monthly request price is $0.20 per one million requests and the free tier provides one million requests per month.
Traditional Virtual Machine (VM) deployment was used for the frontend-tier. On one hand, considering the traffic and budget of the project. On the other, the Machine-Learning (ML) Model required a host for automation and the VM can be reused for this purpose.