Inspiration
YouTube creators often struggle to identify which specific topics and keywords drive high visibility and sustained engagement. Analyzing trending data across multiple regions is a complex and time-consuming manual task. We wanted to build a tool that empowers creators to make data-driven decisions by consolidating global datasets and revealing the key performance drivers—such as views, likes, and keyword popularity—in one place.
What it does
This app is a Streamlit-based dashboard that ingests YouTube Trending CSV files and automatically merges them into a unified dataset. It allows users to filter data by country, category, channel, and date range while providing high-level metrics like distinct video counts and total engagement. Most importantly, it performs a deep-dive Keyword Analysis, extracting terms from titles and tags to compute their correlation with view counts, helping creators optimize their metadata for maximum reach.
How we built it
The application was built using Python and the Streamlit framework for a responsive web interface. We used Pandas for heavy-duty data engineering: converting non-standard trending_date formats to ISO standards, inferring country metadata from filenames, and mapping numeric category IDs to human-readable names. For data visualization, we implemented interactive charts to rank top-performing content and visualize category distributions.
Challenges we ran into
One of the primary challenges was standardizing the schema across multiple regional CSV files, particularly handling the unique YY.DD.MM date format. We also had to optimize the text processing logic to efficiently extract thousands of keywords from tags and titles while maintaining app performance. Defining a precise logic for "distinct videos" was also crucial to ensure that re-trending videos didn't skew the overall engagement analytics.
Accomplishments that we're proud of
We are particularly proud of the Keyword Correlation feature. Instead of just listing popular words, the app calculates how strongly a keyword's presence relates to actual view growth. This provides actionable intelligence rather than just raw data. We also succeeded in creating a "no-code" experience where a creator can simply drag and drop raw data files to receive professional-grade analytics instantly.
What we learned
This project reinforced the importance of data cleaning—specifically how much "pre-processing" is required to turn raw, noisy data into a usable product. We gained deep insights into text mining and statistical analysis, learning how to derive quantitative metrics from qualitative data like video tags. We also learned how to balance comprehensive analytical depth with a clean, user-friendly UI.
What's next for Theme 1 Datahack
The next step is to integrate Sentiment Analysis by processing comment data to gauge audience reaction beyond just likes and views. We also plan to move from static CSV uploads to a Live API Integration for real-time trend tracking. Finally, we aim to implement a predictive machine learning model that suggests the "best time to post" and "ideal title length" based on historical trending patterns.
Log in or sign up for Devpost to join the conversation.