This project analyzes air quality monitoring sensor data to identify cost-optimization opportunities. Specifically, I assess the redundancy of the no (Nitric Oxide) sensor and determine if its removal would maintain data integrity while reducing costs.
My goal is to evaluate whether the no sensor can be eliminated without losing crucial data. This is achieved through statistical correlation analysis and linear regression modeling, using no, no₂ (Nitrogen Dioxide), and noₓ (Nitrogen Oxides) sensor data.
Eliminating the no sensor (Nitric Oxide) in N. Mai, Los Angeles California (CA), will have a minimal impact on overall air quality monitoring. This is based on the strong correlation, interdependence, or redundancy of no with other related pollutants, such as no2 and nox. By leveraging data from these sensors, it can effectively derive no levels, thereby optimally reducing project expenses while maintaining the integrity of air quality data.
- Data Preparation
- Descriptive Statistics Analysis
- Scatter Plot Analysis
- Correlation Analysis
- Line Plot & Time Trends
- Linear Regression Modeling
| Parameter | Count | Mean | Std Dev | Min | 25th Percentile | Median | 75th Percentile | Max |
|---|---|---|---|---|---|---|---|---|
so2 |
46,170 | 2.1603e-4 | 4.0413e-4 | -0.001 | 0.0 | 0.0 | 2.0e-4 | 0.01 |
co |
41,529 | 0.3931 | 0.2528 | 0.0 | 0.2 | 0.3 | 0.5 | 2.0 |
nox |
14,004 | 0.0221 | 0.0206 | 8.0e-4 | 0.0079 | 0.0144 | 0.0288 | 0.1622 |
o3 |
46,634 | 0.0250 | 0.0184 | 0.0 | 0.008 | 0.025 | 0.038 | 0.138 |
pm10 |
46,675 | 29.8535 | 16.2686 | -4.0 | 19.0 | 28.0 | 38.0 | 588.0 |
no2 |
46,689 | 0.0175 | 0.0113 | 6.0e-4 | 0.008 | 0.0147 | 0.025 | 0.08 |
no |
14,008 | 0.0065 | 0.0125 | -9.0e-4 | 3.0e-4 | 0.0013 | 0.006 | 0.1229 |
pm25 |
46,089 | 13.8518 | 9.2207 | -3.8 | 8.0 | 12.0 | 17.4 | 508.0 |
- The
nosensor only has 14,008 data points, which is significantly fewer than others. - The average value of
nois very low, suggesting a minor environmental impact. nohas similar patterns tono2andnox, indicating redundancy.
no(Nitric Oxide) - one of the lowest average concentration, suggests that it might have less effect on the environment.no2(Nitrogen Dioxide) - Linked to combustion or burning processes (like automobiles, industries), higher concentration than ‘no’.nox(Nitrogen Oxides) - total of no and no2, it has a larger value than ‘no’ but is still very low overall.
- Each bar's height indicates the count.
- Majority of measurements for
nocluster is just a small percentage in the range.
- Between 2022 and 2023, there is a noticeable gap for every metric, a sign of missing data or sensor failure.
noxhas seen a significant increase in results following 2023, indicating either changes to the environment, mistakes in measurement, or adjustments to the data collection process.- Possibly related to environmental cycles like temperature or human activity,
no2shows consistent seasonal or periodic behavior. - Changes in the real world, including increased industrial activity, traffic, or changing seasons in pollutant concentrations, may be the cause of the
no2moves and thenoxincrease. - Data gap can be sensors were not working or when problem gathering the data.
- Correlations - the patterns in
no2andnoxmay be connected.
-
[1]
novsno2- There is some correlation between
noandno2, according to the data points. Little disorganized connection. - Not entirely dependent on one another.
- There is some correlation between
-
[2]
novs.nox- Tight rising trend, strong positive correlation between
noandnox. - ‘nox’ is sum of
noandno2, this implies that ‘no’ levels significantly contribute tonoxlevels.
- Tight rising trend, strong positive correlation between
-
[3]
no2vs.noxno2andnoxhave a high positive correlation, just like [2].- This shows how
no2andnoxare interdependent. Changes inno2result innox.
- A linear regression model was built using
no2andnoxto predictno. - The model's results show that
novalues can be precisely predicted from the other two variables.
- R²:
0.9998760560412906 - RMSE:
0.00013389460874487974 - Coefficients:
[-1.0005, 1.0027] - Intercept:
-2.6839e-05
These metrics strongly indicate that the no sensor adds no unique value.
✅ Eliminating the no sensor does not impact air quality monitoring performance.
✅ The regression model can accurately estimate no using no2 and nox.
✅ This results in cost savings for maintenance, hardware, or calibration.
✅ The approach demonstrates the power of correlation analysis and predictive modeling in sensor optimization.
- 📄 Project Selection Hypothesis
- 📄 Methodology Report
- 📄 Results & Discussion
- 📊 Project Cost Optimization Notebook
- Davda, K. (2024, June 27). What is low-cost air quality monitoring, and what are its Working principles? Oizom. https://oizom.com/what-is-low-cost-air-quality-monitoring/
- DD-Scientific. (n.d.). GS+7NO Nitric Oxide (NO) Sensor | Industrial specification. https://ddscientific.com/products/gs-7no-electrochemical-sensor-nitric-oxide-no
- Great Basin Unified Air Pollution Control District (n.d.). Low-Cost Air Quality Sensors. https://www.gbuapcd.org/AirMonitoringData/LowCostSensors/
- Kang, Y., Aye, L., Ngo, T. D., & Zhou, J. (2021). Performance evaluation of low-cost air quality sensors: A review. The Science of the Total Environment, 818, 151769–151769. https://doi.org/10.1016/j.scitotenv.2021.151769
- Kunak Technologies S.L. (2023, June 30). The power of low-cost air quality sensors for cleaner environments. Kunak. https://kunakair.com/low-cost-air-quality-sensors/
- OpenAQ Location ID 7936. (n.d.). OpenAQ Explorer. https://explore.openaq.org/locations/7936
- World Meteorological Organization. (2024, June 13). Low-cost sensors can improve air quality monitoring and people’s health. https://wmo.int/media/news/low-cost-sensors-can-improve-air-quality-monitoring-and-peoples-health