1.2. Seasonal Virus
Influenza is commonly known as the "flu" and is an acute
respiratory disease caused by the influenza virus. Influenza is a highly
contagious disease that causes large and small trends around the world every
year, and 10 to 20 percent of the population usually gets infected within two
to three weeks of its onset [9]. The most obvious symptoms are sudden high
fever of 38 to 40 degrees Celsius within 24 hours of infection, with systemic
symptoms such as headaches, muscle aches and fatigue, and respiratory symptoms
such as sore throat, cough, gabbing, and rhinitis. Dynamically, influenza
usually occurs in late fall and early spring when winter begins [10]. In the
case of influenza, mainly worldwide, it is seasonal influenza, with 5-15
percent of the population infected each year.
1.3. 2019 Novel Coronavirus
SARS-CoV-2 is spread by human-to-human transmission via droplets
or direct contact, and infection has been estimated to have mean incubation
period of 6.4 days and basic reproduction number of 2.24~3.58 person [9]. Many
highly infectious diseases, including COVID-19, exhibit seasonal patterns. For
example, for SARS-CoV, MERS-CoV and Influenza, humidity and temperature are
known as factors of virus survival. By mid-February 2020, the majority of
confirmed cases had occurred in cities between 30 and 50 degrees latitude.
These cities also have similar climatic characteristics with an average
temperature of 5 to 11 degrees and humidity of 47 to 79 percent [2].
2. WEATHER PREDICTION MODEL
In January 2020, Google succeeded in developing a model called
Nowcast, which uses artificial intelligence technology to predict weather
conditions such as precipitation. Nowcast will shorten the analysis work, which
used to take hours, by 5-10 minutes and forecast the
3. APPLIED TECHNOLOGIES
3.1. Hadoop
Apache Hadoop is a Java-based open source framework that can
store and process big data. The two large components of Hadoop were created to
better handle large amounts of data through parallel distributed storage(HDFS)
and distributed processing(MapReduce) [11]. Weather sensors that collect data
every hour across the globe collect large amounts of log data, which are joined
to data analysis using MapReduce because they are semi-structured and
record-oriented. Because there are tens of thousands of weather stations, the
entire dataset has a large number of relatively small files. Due to the nature
of Hadoop, a small number of large files are easy to process and efficient
[5].
- 6 -
weather up to six hours later. Google researchers compared the
new weather forecast model to the existing one based on past weather data from
2017 to 2019. The results showed that new and existing models showed similar
levels of performance in terms of accuracy [3].
2.1. Numerical Weather Prediction (NWP)
Synoptic weather predictions (computer models based on physical
equations) that predict weather from a minimum number of days to a maximum of
two weeks have been developed by the Numerical Weather Prediction. These models
remain the backbone of all weather forecasts. However, this approach has a
critical limitation that will take a considerable amount of time to process the
data [12].
| 