Analysis of the Cryptocurrency ATMs across United States counties to find correlation with different features

Interactive Tableau Dashboard for Visualization

Project Overview:

This project aims to analyze the drives of Crypto ATM adoption across the US counties. We are ultimately trying to answer what factors in particular could potentially influence the adoption of Crypto ATM for each county.

Some of the Potential drivers:

• Unbanking Rate

• Crime Rate

• Population

For this Analysis our Dependent variable is whether Cryptocurrency ATM was launched in a county. Independent Variable are the different drivers and the control variables are State policy, county-level population, characteristics (GDP, density, poverty, diversity, conservative, individualism, and so on)

Potential Methods:

• Lasso Regression Model

• Ridge Regression

Data Gathering:

The data gathering stage involved webscraping and automating large volumes of download mainly using Selenium

Data Cleaning/Merging

After gathering the data large part of the work involved cleaning the data using a mix of Alteryx for automation and Python(pandas).

Exploratory Data Analysis(EDA)

After conducting an exploratory data analysis and running a correlation analysis, it showed that the Total Crypto Currency ATMs across the US counties seemed to be highly correlated with the factors below:

• Violent Crime Rate

• Annual Average Violent Crimes

• Population

• Deaths

• Large Central Metro

• Some College

• Asian

• Hispanic

AND

It also seemed negatively correlated with

• Adults with obesity

• Physically inactive

• 65 and Over

MAPPING THE DATA using PLOTLY (Chlorpeth)

On a quick glance there seems to be an observable pattern between these two charts.

Conclusions

So far analysis has indicated that there seems to be a strong correlation between Crypto ATMs with

• Annual Average Violent Crimes

• Population

• Deaths

Also, it is interesting to observe that counties with higher Asian demographic seem to have more correlation with the Crypto ATMs, however, this needs further analysis with non Hispanic data and other demographic data What we also observed was that there was a very strong correlation with unbanking rate and crypto ATMs, also the unbanking was highly correlated with Annual Average Violent crimes, deaths and the population of the county.

Moving forward, this finding has opened a new dynamics for research into these county health measures. The next stages of this project underway are identifying important features and adding more variables such as county ideology and individualism to the dataset.

CORRELATION HEATMAP GENERATED IN PYTHON