Multisensor Dataset for Alcohol Presence Detection (MQ-3, MQ-135, DHT22)
Description
This dataset has 1,942 rows of sensor data collected to detect alcohol using MQ-3 and MQ-135 gas sensors and a DHT22 sensor for temperature and humidity. Data were gathered in both lab and indoor environments at East West University, Dhaka (23.7833°N, 90.3833°E). The MQ-3 sensor is sensitive to ethanol, while MQ-135 detects various gases. The DHT22 sensor measures the surrounding temperature and humidity. Alcoholic samples had ethanol levels between 5% and 30%, and non-alcoholic samples included water, tea, coffee, and soft drinks. A fixed process was followed for each test. Sensors were preheated for 24 to 48 hours. Each test began with a 10-second waiting time, followed by sensor readings every 30 seconds for 2.5 minutes. A clean air flush was done between samples to clear leftover gas. Each row in the dataset has values from MQ-3, MQ-135, temperature, humidity, and a label showing if alcohol was present (1) or not (0). Tests show that sensor values from MQ-3 and MQ-135 clearly differ when alcohol is present, while temperature and humidity do not show much change. Graphs such as box plots and PCA also show that the two classes are easy to tell apart based on sensor data. The dataset is balanced with 961 alcohol and 981 non-alcohol samples. It is useful for training machine learning models for alcohol detection in cars, workplaces, and portable devices. No people or animals were involved, and data collection followed research ethics. Please cite the dataset using the following DOI if you use it in your work: 10.17632/kvcfp5y95z.1.
Files
Steps to reproduce
The data was collected using MQ-3 and MQ-135 gas sensors along with the DHT22 sensor for temperature and humidity. The setup was placed in both controlled lab and indoor environments at East West University, Dhaka. Sensors were first preheated for 24 to 48 hours to stabilize their readings. Alcoholic samples were prepared using ethanol solutions ranging from 5% to 30%. Non-alcoholic samples included water, tea, coffee, and soft drinks. During each test, the sample was placed near the sensor, followed by a 10-second waiting time to let the sensor adjust. Sensor readings were then recorded every 30 seconds over a period of 2.5 minutes. After each sample, a 5-minute clean air flush was performed to remove any leftover gases. All readings were collected using Arduino-based hardware and logged to a computer in real time. Each row includes the analog values from MQ-3 and MQ-135, temperature and humidity from DHT22, and a binary label showing the presence (1) or absence (0) of alcohol. This method was repeated for each sample to ensure consistency and reliability.
Institutions
- East West University