The soundscape of global music is changing, faster than ever. TikTok virality, global genre crossovers, and streaming algorithms are reshaping what we listen to and why. But what if we could predict where the cultural current is headed next?
This project dives into that question by forecasting monthly search interest (2018–2024) for 9 major music genres and projecting their trajectories through 2025 and 2026. Think of it as a data crystal ball for music execs, brand strategists, and fans alike — spotlighting what genres are poised to surge, fade, or surprise us next.
- Label execs & A&R scouts looking for the next big sound
- Brand collab teams identifying emerging music movements
- Streaming platforms shaping recommendation algorithms
- Artists curious about genre momentum & timing their releases
I used Google Trends data, a real-time signal of collective curiosity, because it’s not about what people say they like, but what they actually look up. It’s raw, reactive, and highly telling.
To collect the data:
- I used PyTrends, the unofficial Google Trends API, pulling monthly interest scores from January 2018 to December 2024.
- Due to API limitations (5 terms max), I divided the 9 genres into two batches and merged them for continuity.
- Final dataset: 9 genres × 84 months = 756 signals of shifting cultural taste.
- The resulting dataset contains monthly interest scores (0–100) for each genre.
- Columns represent genres, rows represent months from 2018 to 2024.
- Cleaned to remove low-variance columns (e.g., R&B was dropped due to insufficient variation).
genres = ['hip hop music', 'pop music', 'r and b', 'country music', 'rock music',
'alternative music', 'kpop music', 'metal music', 'latin music', 'indie music']Genres like R&B were removed due to low signal variance, a story in itself about fading public interest or the term not suitable enough to collect trend data.
Music trends have rhythm, both metaphorically and statistically. That’s why I didn’t just pick any model, I used SARIMA, a time series model that embraces seasonality. We’re not just predicting more of the same, we’re mapping the when and why behind peaks and dips.
Each genre went through the same modeling workflow:
- Stationarity checks via ADF tests.
- Seasonality analysis to detect yearly rhythms.
- Train/test splitting for real-world validation.
- Auto-tuned SARIMA modeling (with manual diagnostics to avoid blind automation).
- Forecast generation and visual storytelling.
- Accuracy scoring and fine-tuning.
If you're more interested in the story the data tells, not the modeling journey, you can jump straight into the final visualizations, forecasts, and genre insights.
Explore the historical and predicted trends for each genre, backed by cultural context and data visualizations:
👉 Click here to explore the results.
You'll find:
- 2-year forecasts for each genre
- Clear trends, turning points, and seasonality patterns
- Commentary on what the numbers reveal about music, culture, and what's next
👉 Check out the full notebook here.
This walks through each genre's stationarity, seasonality, model tuning, evaluation scores, and 24-month forecasts, everything from raw signal to cultural readout.
All code is written in Python, using:
- pmdarima.auto_arima for parameter selection
- SARIMAX from
statsmodelsfor modeling - Matplotlib for storytelling through visuals
- MAE, RMSE, MAPE for evaluation
- Manual diagnostics like AIC, BIC, Ljung-Box and Jarque-Bera to ensure robust results
Here’s a teaser of what that looks like for Hip-Hop:
To build reliable forecasts using ARIMA or SARIMA, the data needs to be stationary, meaning its statistical properties (like mean and variance) don't shift over time.
I ran an Augmented Dickey-Fuller (ADF) test to check this. The result?
p-value = 0.296 (way above the 0.05 threshold)
So, no — Hip Hop's trend is non-stationary, which means we’ll need differencing in our model (hence the "I" in ARIMA).
Yes, but it’s weak.
I performed a seasonal decomposition and saw a repeating yearly pattern, although not very strong. The seasonal strength score came out to 0.417 (closer to 0 = weak). Here's what the breakdown showed:
- Trend: A consistent downward trend from 2018 to 2022, then it flattens.
- Seasonality: Some repetition every year, but not super pronounced.
- Residuals: Random fluctuations, fairly balanced around zero, which is a good sign.
This justified the use of SARIMA (which accounts for both seasonality and non-stationarity) over simpler ARIMA.
👉 Learn more about Seasonality.
To simulate future forecasting conditions:
- I trained the model on all data except the last 12 months
- I tested the forecast performance on the final year
This split mimics how forecasts are used in real-world settings, we test predictions on unseen data.
Instead of guessing the (p, d, q) and seasonal values, I used auto_arima() to find the best-fitting parameters — but with a twist.
Before running it, I already:
- Checked for stationarity (to guide the value of
d) - Measured seasonal strength (to decide if seasonal terms were needed)
So this wasn't blind automation — it was smart automation, guided by the nature of the data.
Final model: SARIMA(1,1,1)(1,0,1)[12]
Using the selected SARIMA model, I forecasted the next 12 months and compared predictions to the actual test data.
Evaluation metrics:
- MAE: 0.36 → average error size
- RMSE: 0.42 → no large error spikes
- MAPE: 7.5% → excellent accuracy (below 10% is strong)
The model also passed residual diagnostics:
- No autocorrelation (Ljung-Box p = 0.82)
- Residuals are normally distributed (JB p = 0.53)
- No heteroskedasticity (H p = 0.73)
Visually, the forecast hugged the real values closely.
With the model validated, I extended the forecast by 24 months to cover 2025–2026.
Key takeaways from the final forecast:
- The downward trend continues, but slows down
- Seasonal ups and downs persist — possibly tied to cultural/industry cycles
- Hip Hop popularity may be stabilizing, not crashing
Even Hip-Hop — a once-undisputed leader — shows a downward trend post-2021. Not a collapse, but a flattening curve... maybe making space for genres like Latin, K-Pop, or Indie to step up?
👉 View the forecast results for all genres here.
The graph displays forecasted popularity trends for several music genres from March 2025 to November 2026.
-
Metal shows the most dramatic and consistent upward trend. Starting around 40, it rises steadily to become one of the most popular genres, reaching nearly 90 by November 2026. This indicates a strong surge in popularity.
-
Alternative exhibits continuous and strong growth, moving from around 40 to reach popularity in the high 80s by the end of the forecast. It is consistently gaining popularity.
-
Latin, after an initial dip, it shows consistent and significant growth from late 2025, climbing from its starting point of around 35 to the high 70s by November 2026.
-
Kpop starts relatively high and sees a slight dip in mid-2025, but then generally trends upwards to reach the high 70s by November 2026, indicating moderate overall growth.
-
Country genre's popularity at the end of the period is still respectable (around 40-50), but its path is very bumpy.
-
Rock experiences an initial decline but then shows a strong resurgence in early 2026, peaking around 60. It ends the period in the high 30s, meaning it recovers from its low but doesn't necessarily achieve consistent growth above its starting point.
-
Hip Hop shows a consistent and significant downward trend. Starting as one of the most popular genres, it drops sharply and steadily, ending in the low 20s by November 2026.
-
Pop genre, while starting as the most popular, it undergoes a sharp initial decline and then stabilizes at a much lower popularity level (around 50-60) for the remainder of the forecast. It doesn't recover its initial dominance.
-
Indie experiences a very significant and sustained decline throughout 2025, plummeting from its relatively high starting point of 70 to lows around 10-20. It shows only a slight recovery in 2026, ending as one of the least popular genres.
Strong Growth: Metal, Alternative, Rock.
Stable/Fluctuating: Latin.
Declining/Volatile: Hip Hop, Pop, Country, Kpop, Indie. (Note: Some of these, like Pop and Rock, have strong recoveries after initial dips, but their overall trajectory might be seen as more volatile than consistently growing).
genre_trends_forecasts.md: File with plots and forecast outputsConcepts.md: In-depth breakdown of stationarity, seasonality, and SARIMAdata_retrieval.ipynb: Code for Google Trends data retrievalforecasting.ipynb: Full SARIMA modeling notebook (all genres)visualizations.ipynb: Comparison plots for all genresforecasts.csv: Forecast time series datatrend_df.csv: Cleaned Google Trends dataset (2018–2024)normalized_df.csv: Normalized forecast time series data
This isn’t just data. It’s a mirror of culture, of what moves us, what fades, and what pulses just beneath the surface. Music isn’t static, and neither should our understanding of it be.