Home > Article > How CLM Machine Learning Works as Covid-19 Self-Assessment Tool
Aditya Gagat Hanggara
28 July 2020
In the early days of computer technology, a programmer must write lines of code and algorithms just to do a certain task. However, with the increasing variety of functions that modern computers - and even mobile phones - now can do, such traditional methods are becoming less efficient and not practical. Especially if the task requires a large data set and regularly gets updated. This is the reason why machine learning-based technology was adopted.
Machine learning is a subset of Artificial Intelligence, and as the name suggests, puts the focus on learning carried out by the machine/computer itself. Even so, human involvement is still needed. However, with this concept, programmers no longer need to go through complicated and convoluted processes by writing code explicitly one by one.
With just one calculation model, machine learning allows computers to optimize results by learning from the given dataset or past experience. The model created can be predictive to produce a prediction in the future, or descriptive to obtain knowledge from studied data.
Now, can you name a simple example of a machine learning application that you know? No need to see far away Smartcitizen, because you just have to check your email address. You know about the spam filter feature, right? Yes, this feature is one of the early examples in the application of various machine learning techniques. This method is used by email service providers such as Gmail and Yahoo Mail to identify spam emails by analyzing similar emails on a computer.
With the development of the corona cases in the capital, the Government of Jakarta continues to take steps to mitigate the rate of spread of the Covid-19 pandemic. In its implementation, the provincial government also utilizes information technology systems such as the Corona Likelihood Metric or CLM which were built through collaboration with the Harvard CLM Team and Klakklik.ID. The risk test tool is then embedded in the JAKI application as the latest feature, now also known as JakCLM. It. can be used by citizens who domiciled or are doing activities in Jakarta to calculate the risk of contracting the coronavirus.
At a glance, the appearance of CLM is not much different from the self-screening tools that had already been circulating in the general public. However, CLM currently is the only tool that has the advantage of machine learning technology. This way, CLM can process the answers given by the user and then provide the results of the test along with recommendations that suit user conditions, including recommendations for PCR tests if needed.
Bahrul Ilmi Nasution, a data volunteer who is also a graduate of the STIS Statistics Polytechnic and is now a member of the Harvard CLM Team, explained that the use of machine learning technology has enabled CLM to provide test results with high sensitivity.
"By using machine learning technology, CLM has advantages in terms of scalability. In other words, CLM can be used by the whole community without worrying about costs or limited stock. Also, the CLM was developed using Jakarta patient data, so it has good accuracy and sensitivity values," he said.
In processing the answers given by the user, CLM uses three types of data as a reference: demographic conditions (age, sex), history of symptoms of Covid-19 in the last 14 days, as well as patient history (travel, contact, and health workers).
In addition, CLM machine learning models have also been trained with patient health history data that has been examined by the DKI Jakarta Health Office and identified Covid-19 to improve accuracy.
"So, at first, the person filled out the questionnaire in the JakCLM application. After that, the required data is sent to the CLM application programming interface (API) to be processed using the machine learning model," explained Bahrul about the processing of CLM user answers.
"This machine learning model has been trained based on the data of thousands of other Covid-19 patients. When the machine is given new user data, the machine will then give a prediction of 'how likely it is that this user has Covid-19' in the form of a percentage of 0-100%. This value will be converted into three risk categories: low, medium, and high."
During the process of making the CLM machine learning model, Bahrul was not working alone. He collaborated with Jessica Audrey Wijaya, a Harvard University student who is currently pursuing a Master of Science in Data Science program.
[Jakarta X Harvard CLM Team: A COVID-19 Self-Checker Development Collaboration]
Bahrul admits that machine learning technology was still quite foreign to some people. For example, in the first few days of JakCLM operation, a number of users asked why they were receiving ‘not safe’ status, even though they did not feel the symptoms of Covid-19.
Responding to these questions, Bahrul said: "The possibility of Covid-19 a 20-year-old young man who has/has no symptoms is very different from a 70-year-old grandmother who has/does not have the same symptoms. Another example, the possibility of Covid-19 asymptomatic youth will be different from an asymptomatic health worker. A health worker who treats patients in a hospital basically has a higher risk. Therefore, we must look at the user data thoroughly, not just specific symptoms.
"Secondly, the scores generated by the CLM model are also relative. There are no right and wrong answers in categorizing a score into a category of low risk (safe), moderate risk, or high risk. Previously, the process of converting scores into our three categories (low, medium, high risk) can be considered very strict and conservative. After further discussion, we decided to relax this process by raising the threshold of low and moderate risks, which should have calmed the concerns of the people.”
Smartcitizen, now you know that CLM machine learning technology is very helpful in calculating the risk of you contracted with Covid-19. So don't forget to check your symptoms on CLM regularly. Also invite your friends, relatives, and close family members to use it. While doing so, let's also continue to suppress the spread of the corona pandemic, by routinely applying daily health protocol: Washing hands, Wearing a mask, Maintaining distance!
Aditya Gagat adalah lulusan Teknik Informatika dari Binus University yang saat ini menjadi salah satu Content Writer di Jakarta Smart City. Gemar mengamati isu transportasi, olahraga, teknologi dan sains, ia memulai karier Jurnalistik bersama media internasional Motorsport.com pada 2016-2019. Saat ini ia terfokus pada topik kesehatan, khususnya mengenai penanggulangan pandemi Covid-19 di wilayah DKI Jakarta.
04 June 2021
09 May 2020
22 July 2021
26 August 2021
05 August 2021