Music Industry: Taylor’s Era
Completed Nov. ‘23 - Jan. ‘24
(CareerFoundry Capstone Project)
-
This dashboard was a culmination project for the CareerFoundry program and therefore meant to showcase technical skills for mentors and fellow students, while also providing an introduction to Taylor Swift’s music. Future projects will have a more focused audience, specifically to non-technical audiences and music industry experts.
-
Google Search Scores (Search popularity across album releases from Google Trends)
Spotify Audio Features (Source: Kaggle user)
Billboard Hot 100 Songs & Top 200 Album Charts (1958-2023, GitHub user)
Week 1 Album Sales (2007-2023, manual collection)
Total Spotify Plays (Source: Spotify, manually imputed)
-
Domain-specific research, audience diversification, and creative technical data wrangling
-
Open-source data
Data Quality Reporting
Creative data wrangling
Exploratory data analysis (correlation, pair plots, heatmaps)
Python visualizations
Geospatial analysis
Supervised Machine Learning: Linear Regression
Unsupervised Machine Learning: Principle Components Analysis, K-means Clustering
Time- Series Analysis and Forecasting (ARIMA)
Differentiated Dashboard Storytelling
-
Python, Jupyter Notebook, Tableau, Excel
-
Pandas, Seaborn, NumPy, Matplotlib, Folium, scikit-Learn
This project was a capstone project for the CareerFoundry program with a wide purpose and project benchmarks that did not include accessibility. I am independently learning about making data analytics accessible. All people have the right to access data visualizations and content. I do not believe that the final dashboard is fully accessible. Consequently, I am working on more focused projects tied to this data as well as improving accessibility for all past and future projects, including those about Taylor.
Context and Objective:
This analysis examines success measures and sound characteristics of Taylor's music, showcasing the analytical process. “Taylor’s Era” was a capstone project for the CareerFoundry program, however the exploratory analysis also seeks to launch a larger project about Taylor’s music, targeted at music fans and non-technical audiences.
Domain-specific research and open-source data collection fueled an exploratory data analysis to answer key questions. Creative wrangling was essential to relate all metrics and investigate relationships.
The Process
I searched for correlations between album and song success measures as well as audio features by creating pairplots and correlation heatmaps to visualize potential relationships. I conducted regression analysis to identify if there was a linear relationship between focus variables to test each hypothesis using Supervised Machine Learning. (Python)
My data had many variables, so I used Principle Components Analysis to reduce dimensionality and implement K-means. K-means clustering resulted in 4 distinct album clusters and song clusters, each with unique patterns across success measures. (Python)
I showcased each album cluster's performance in success measures and audio feature characteristics.
I conducted a time-series analysis using Billboard Hot 100 Song Charts. I decomposed each time series using an additive model, tested for stationarity with the Dickey-Fuller Test, and stationarized the data through iterations of smoothing data with differencing. After testing for autocorrelation, I used the ARIMA model to attempt to forecast future song chart success, but the model was not a good fit due to residual shape and external variable fluctuations. Future exploration could include revisiting the time series through other models, such as the GARCH.
I creatively compiled and merged data from Google Trends to conduct geospatial analysis. By tracking periods during each album's release across the United States, I showcased regional trends that show changing fan base, album preferences, and overall search popularity. While Google Trends is an automated analytics tool, this work is one example of the need for people- not just machines- to interpret data in a meaningful way.
I identified album cluster patterns across time for Billboard Album charts.
I examined patterns within the famous re-recording process. All Taylor’s Versions except for Speak Now(TV) have surpassed their originals in total album streams, which is impressive considering the difference in time on Spotify. This was one way to produce data-backed insights to bridge the importance of “Taylor’s Versions” to fans with technical audiences.
Every good analysis develops insights anchored in key questions. I have particular strengths in creating follow up explorations to progress organizational understandings.
This is one example of translating a technical exploratory analysis into a highly interactive dashboard.
