AIlert is a deep neural network that predicts possible hazards for the next few days around the world. Also, this is a mobile application through which you can receive information about hazards in selected locations and regions. This application is ready to notify people about possible disasters nearby, so they could avoid injury or even death.
This challenge is aimed to create automatic detection of hazards around the world. We have created a neural network that makes disaster predictions based on natural problems, not human factors. Specifically, we paid attention to earthquakes and landslides.
This neural network creates forecasts every 24 hours for the following days, so everything you need is to get access to this data. Therefore we have created a mobile application with a map, where you can select the area of interest. However, a person can also be alerted to a threat that will be nearby his location.
We believe that our project will be of use to everyone.
Firstly, everyone can take care of themselves. Secondly, the government can calculate the potential danger zones and take precautions for their country or regions. At last, farmers and businessmen whose activity is connected with growth of grain crops can ensure that their fields are safe from nature threats.
The neural network itself consists of several layers of neurons, so it produces fairly good indicators on rather big data. It processes datasets that have been fetched from NASA and ECAD. Specifically, it considers such measurements as: temperature, pressure, radiation, cloudiness, rain and others.
The data is scientifically reliable, because some sets have been collected by scientists since the 19th century every day for each country, while other data are collected from satellites and posted on the network.
We hope that our application can help people and solve their problems, so we plan to:
- Improve our neural networks
- Improve API
- Improve the work with the application
- Add new types of hazards
- Make more accurate long-time predictions
Before we chose our topic we were inspired to work on the problem that is spread all over the world. This problem is a frequent natural hazards. It is a well-known fact that fire or earthquake might have negative effects on industry and economy, including the environment itself. We strongly believe that we can create something that will predict disasters, so that they could be avoided, in order to remit damage.
Initially, we studied a lot of data provided by NASA, JAXA, ECAD, USGS, NEIC We have selected the appropriate data for our task. Next, we thought about how to prepare the data and process it. Then, we discussed what neural networks or other models can be selected for our datasets to make the predictions as accurate as possible. To demonstrate the forecast results, we have created a mobile application for Android. We have chosen as an interactive a map, on which you need to select an area of interest for further analysis.
We used an integrated approach to create our project. Creation of AI, API, Android App and Landing Page. On the backend, we used such programming languages like:Python, PHP, and for the Frontend part, Android Java. We used Adobe Photoshop, Illustrator, Canva, etc. to create mockups and the designs.
We had problems with predicting all types of hazards, as it takes a lot of time. It also takes much time to improve the accuracy of the neural network forecast. However, we can already say that we have achieved sufficient functionality of the application, because we can select different regions of countries, monitoring all the hazards at once, and filter them.
When we studied the data provided by NASA and their partners, we selected the USGS, CSA datasets on earthquakes and landslides. We also used ECAD datasets to get data on temperatures, radiation, pressure, humidity, precipitation and other measurements. We chose this particular data, because it has great affect on the accuracy and consistency of hazards.
- CSA: ftp://data.asc-csa.gc.ca/users/OpenData_DonneesOuvertes/pub/CASSIPOE/
- NEIC: https://www.kaggle.com/usgs/earthquake-database
- USGS: https://www.usgs.gov/natural-hazards/earthquake-hazards/lists-maps-and-statistics
- ECAD: https://www.ecad.eu/dailydata/predefinedseries.php
(#hazards, #ml, #ai, #neural_networks, #keras, #python, #api, #java, #android, #php, #mobile, #google, #maps, #landslides, #earthquakes, #allert)