Predictive Analytics of Air Alerts in the Russian-Ukrainian War

Starting from February 24, 2022, the date of Russian invasion in Ukraine, it has been very important to understand the structure and patterns of air alerts and predict when and how long an air alert is going to take place. Initially, we created a channel in Telegram Messenger social network for air alerts forecast in Ukraine. The main approach was based on loading messages from other Telegram Messenger channels, analyze them using NLP methods, and then, using experimental heuristics, make prediction when an air alert is about to start. Currently, there are many similar channels in Telegram Messenger, which publish up-to-date information about current air alerts. At the same time, there are many datasets with historical data about air alerts. Our experience of air alerts intuitively shows that there is a geospatial pattern in emerging alerts in different regions of Ukraine. As a result, knowing the cause of the alert and how alerts propagated in different regions, we can anticipate when an air alert is going to start and how long it is going to last in our region. Air alerts analytics is also considered in [1, 2, 3, 4, 5, 6, 7, 8]. The main goal of our study is to conduct an exploratory data analysis and create a predictive model to forecast the duration of air alerts.