Watch: Fan-Out Query Search Volume Prediction Using Deep Learning
A deep learning approach using a Query Demand Estimator to automatically predict search volume ranges for long-tail queries generated by a fan-out model.
Transcript
Finding new search terms for your website is easy with modern language models, but figuring out which ones actually get traffic is a massive bottleneck. That is why we developed the Query Demand Estimator, or QDE.
The QDE is a deep learning model designed to automatically predict search volume ranges for millions of generated keywords. We trained the model using historical search data, grouping queries into twelve distinct volume buckets.
While predicting the exact volume bucket is challenging, our model gets it right, or very close, nearly fifty-five percent of the time. Because it is choosing from twelve possible ranges, this performance is far better than random guessing. It gives content teams a fast, reliable way to prioritize their search engine optimization efforts.
For queries that have never been seen before, the most promising approach is entity tracking. Since most searches are about specific people, products, or brands, we can predict search volume by analyzing how popular those entities are across Wikipedia, knowledge graphs, and news trends.
By combining deep learning with entity analysis, businesses can turn a massive list of raw keywords into a prioritized, actionable roadmap for organic growth.