Name: Air Quality Dataset
Creator: Kaggle
License: https://creativecommons.org/licenses/by-sa/4.0/

About this Dataset

Air Quality Dataset

Context

The variation in indoor gas concentration over time is monitored and data is stored in order to use this information to evaluate the type of activity carried out in the room. Thanks to the use of artificial intelligence, a quantitative approach in determining the gas concentration was avoided, which would have required careful calibration of the sensors. The dataset contains the values acquired by an array of 6 low cost sensors in successive instants of time, and the stored values are associated with the particular action that generated them. Through an appropriate data processing, based on machine learning algorithms, after an initial training phase it is possible to recognize the actions that are carried out inside the home.

Content

The presence of chemicals in the air is determined through a series of electrochemical gas sensors that have been selected based on the stated technical specifications on the ability to detect classes of compounds. The sensor set can be grouped into two main categories:

MQ sensors (MQ2, MQ9, MQ135, MQ137, MQ138) which have great sensitivity, low latency and low cost; each sensor can respond to different gases;
Analog CO2 gas sensor (MG-811) which has excellent sensitivity to carbon dioxide and is scarcely affected by the temperature and humidity of the air.
The dataset contains 1845 collected samples describing 4 target situations:
1 - Normal situation - Activity: clean air, a person sleeping or studying or resting - Samples: 595;
2 - Preparing meals - Activities: cooking meat or pasta, fried vegetables. One or two people in the room, forced air circulation - Samples: 515.
3 - Presence of smoke - Activity: burning paper and wood for a short period of time in a room with closed windows and doors - Example: 195.
4 - Cleaning - Activity: use of spray and liquid detergents with ammonia and / or alcohol. Forced air circulation can be activated or deactivated - Samples: 540.
Each sample is made up of 7 values; the first six values are the sensor outputs, while the last is the index of the action that generated the values acquired by the sensors. The four different situations are associated with a fairly different composition of the air, taking into account that any activity produces chemical substances due, that is, to human respiration, to the exhalations of metabolic processes, to the release of volatiles by combustion and / or oxidation, and evaporation of household detergents.

Acknowledgements

Gambi, Ennio (2020), “Air Quality dataset for ADL classification”, Mendeley Data, V1, doi: 10.17632/kn3x9rz3kd.1

Tables

Dataset

@kaggle.saurabhshahane_adl_classification.dataset

34.93 kB
1,844 rows
7 columns

CREATE TABLE dataset (
  "n_670" BIGINT  -- 670,
  "n_696" BIGINT  -- 696,
  "n_1252" BIGINT  -- 1252,
  "n_1720" BIGINT  -- 1720,
  "n_1321" BIGINT  -- 1321,
  "n_2431" BIGINT  -- 2431,
  "n_4" BIGINT  -- 4
);