Engineering Analytics (ESE1007) – Air Quality data from Network of Sensors, Singapore

Project Title: Air Quality data from Network of Sensors

Background:

In a World Health Organisation (WHO) study, Krakow has been rated among the most polluted city in the world. In the report, Krakow was ranked 8th among 575 cities for levels of PM2.5 and 145th among 1,100 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. With a state-run network of 8 monitoring stations, Krakow has decided to go a step further by building a network of low-cost air quality sensors that can be deployed across the entire city. The first step in the fight against the smog is to identify areas of problem and to raise awareness among residents and the authorities.

Objectives:

The objectives of this project are:

1. To pre-process the data to ensure data is clean and ready for next stage of analysis.
2. To perform data transformation so as to gain insights into the data.
3. To perform data visualization so as to discover patterns, trends and etc.
4. To fit the data into a Linear regression model for real value prediction.

Data Set Information:

The dataset consists of air quality data (the concentrations of particulate matter PM1, PM2.5, PM10, temperature, air pressure and humidity) from 2017 generated by a network of 56 low-cost sensors located in Krakow, Poland. Each sensor has its own location with an identifier id, latitude and longitude recorded in the sensors_location.csv file. The measurements are grouped into 12 files, one for each month of the year. The resolution of the sensor data is 1 hour.  Each record in the data set consists of 337 attributes:

• UTC time
• 3_temperature: temperature at location id 3
• 3_humidity: humidity at location id 3
• 3_pressure: pressure at location id 3
• 3_pm1: PM1 measurement at location id 3
• 3_pm25:5 measurement at location id 3
• 3_pm10: PM10 measurement at location id 3
• 140_temperature: temperature at location id 140
• ….. all the way for all location id

Known Issues:

• PM1 is not calibrated and therefore it can be bigger than PM2.5
• 5 can be bigger than PM10 within the limits of measurement error
• For January and February 2017, the humidity and temperature sensors were not calibrated. Therefore, it may show inaccurate values.