Noise Pollution EDA (Exploratory Data Analysis)
Presented by: Andy K., Sidharth M., Aidan V., Wayland L.
OUR TEAM!
Andy Kuang
Data Science + Economics
Sid M.
Data Science + Intended Haas + Urban Science
Wayland La
Applied Math + Intended Data Science
Aidan Van Dalen
Cognitive Science + Intended Data Science
Why this Project?
Noise pollution is often glanced over compared to other pollutants, such as air or water pollution. Therefore, we decided to explore noise pollution by analyzing a US cities population dataset on Kaggle and a noise pollution dataset on OpenStreetMap (OSM) to find patterns in the relationship between Population Density and corresponding Noise Pollution.
Our ultimate project goal was to create a site where tourists could identify more tranquil areas with less noise near the cities they plan to visit, thus giving users a better vacation experience.
Data Set Overview
We analyzed two datasets on Population Data and Noise Data. The data combined contained 75 rows and 34 distinct columns. Some important columns to note were city, density and N_median (noise median).
We tried finding datasets combining population and noise pollution data but all of the datasets we could find were excessively large—some more than 12 GB.
Data Cleaning
We removed all duplicate cities and combined the datasets using the city column. (Removed duplicate rows and delete unnecessary columns).
Combined the two data sets and removed the outlier in the average noise pollution column.
Heatmap
We developed a correlation matrix heatmap to understand associations between certain variables. We noticed that there was a negative association with population density and pollution. Which prompted us to delve deeper into density vs noise pollution.
The heatmap displays correlation coefficients between different numeric features but we can still see correlations between MAX & RANGE of noise pollution with population and population density.
Hypothesis Testing
Question: Is there a correlation between Population density and excess noise pollution?
Null hypothesis: There is no direct correlation between population density and excess noise and any inconsistencies are due to either menial factors or chance
Alternative hypothesis: The proportion of excess noise to population density are correlated for all cities
The Plan: Linear Regression + Bootstrapping
Observed Value (Slope of Linear Regression): 0.0115644
Testing Results
Tracking Slopes
Histogram of Slopes
Conclusion: We can reject our null hypothesis and say that there is a correlation between the two variables due to the absence of zero as a possible slope in histogram
Density Map_box
Through density mapping, we can see which cities in the U.S are highly congested.
This is one of our most useful graphs as we believe density could be directly correlated to noise pollution.
E.g. code
Map of Noise Pollution: The high density areas and the areas with higher concentrations of noise pollution are similar.
Since San Francisco, Boston, NYC, and LA are very dense in comparison to the other cities, the density map is realistic.
Population Density
Noise
Choropleth Mapping
Choropleth Mapping allows for visualizing geospatial patterns.
For this project, it helps us understand how total population distribution correlates with noise pollution.
E.g. code
Mean Noise Pollution by City/State in the U.S. You can see how lack of data can affect the output of Decibels of Noise Pollution as somehow California has a lower mean than Kansas.
Population Density by City/State in the U.S. You can see possible similarities between Pop Density & Db, but also outliers like Illinois.
Population Density
Mean Noise
Random Forest Regression
Random Forest Regression
We used Random Forest Regression to fill in a predicted_noise_pollution value for every city in Populations dataframe. Lack of proper data caused values to be too similar. If we had more cities, the predicted values would have been more accurate.
average squared difference between the actual and predicted values
Conclusions
We got in touch with Lukas Martinelli to ask if we may use his API on Global Noise Pollution. Regretfully, he informed us that since we weren't working with a company, he wouldn't be able to provide the key.
Noise data:
Population data:
This noise data sample was not large enough make generalizations about the noise data in all the cities in of our population data.
We plan to implement a search bar dropdown menu that will allow users to choose any US city of their choice and provide them with noise pollution information.
Noise Dataset Difficulty
Noise Data
Dropdown Menu
THANKS!
Especially Madhuri & Alan!