pytorch实现RNN实验.rar

# Context In the past, the ancient city of Krakow was known as the capital of Poland. In 2000, it became known as the official European Capital of Culture. Now, it is known for having some of the most polluted air in Europe… In a World Health Organiation (WHO) study Krakow has been rated amongst the most polluted in the world. In the report, city was ranked 8th among 575 cities for levels of PM2.5 and 145th among 1100 cities for levels of PM10. Hazardous air quality is a common problem particularly during the colder months when many residents use solid fuels (mostly coal) for household heating. Air pollution in Krakow poses a significant danger to human health and life. Krakow's poisoned air includes amongst other things: particulate matter, benzo(a)pyrene and nitrogen dioxide. The state-run network of monitoring stations consists of 8 monitoring stations in Krakow. We decided to go step further - to build network of low-cost air quality sensors that can be deployed across entire city. The first step in the fight against smog is to identify areas of problem and to raise awareness among residents and the authorities. It is very important to create a network of sensors – only then you can check the actual conditions in various areas of the city. The technology enables real-time monitoring of air quality via map.airly.eu, so the information about the air in a specific location is easily accessible and always up to date. # Content This dataset consists air quality data (the concentrations of particulate matter PM1, PM2.5 and PM10, temperature, air pressure and humidity) from 2017 generated by network of 56 low-cost sensors located in Krakow, Poland. Each had its own location (6 of them where replaced during this time period and have almost the same latitude and longitude). Measurements are grouped in 12 files, one for each month. Resolution of data is 1 hour. # Known issues: - PM1 is not calibrated and therefore can be bigger than PM2.5 - PM2.5 can be bigger than PM10 within the limits of measurement error - for the first two months humidity and temperature were not calibrated and therefore can show inaccurate values # Acknowledgements The data was generated by Airly network - the project is still in its beginning stage, but over 1000 sensors have already been implemented in Poland. Airly is a startup definitely worth watching, especially for citizens of the most polluted cities. After all, it’s clean air we all want to breathe. # Inspiration I think that this dataset offers some great opportunities for predictive models and data visualization. Airly's goal is to develop an effective forecast and monitoring of air quality, employ Artificial Intelligence and utilise data from extensive sensor network. If anyone has any ideas, breakthroughs or other interesting models please post them. # Some questions worth exploring: - What are the best prediction models based on extensive sensor network - statistical or numerical forecast? - How weather affects air quality? - How much pollution comes from cars, factories and coal-fired power plants?