About

The most useful insight from my project was that search volume was the strongest predictor of trend growth. After training the machine learning models on 585 Google Trends records and engineering 14 features, I found that search volume alone contributed 99.67% of the model's prediction power, while the other 13 features—such as sentiment, language, text length, and time of day—together contributed only 0.33%.

The regression model achieved an R² score of 99.98% with an RMSE of 0.0008, and the classification model achieved 100% accuracy in predicting slow, medium, and fast-growing trends. These results showed that trend velocity is primarily determined by search volume and the duration for which the trend remains active.

Based on this insight, I recommended that businesses focus on monitoring search volume spikes and trend duration instead of relying heavily on sentiment analysis. This helps identify emerging trends much earlier and supports better decision-making for content creation, marketing campaigns, and real-time business intelligence. The project demonstrated that a simple, data-driven approach based on search volume can accurately predict which topics are likely to gain rapid public attention.