# TrueRoute | Sherwin Vishesh Jathanna

> TrueRoute analyzes historical flight data, weather patterns, and passenger sentiment to predict reliability. It assigns a score to flights based on delay risks…

Source: https://www.sherwinvishesh.com/project/trueroute  
Copyright 2026 Sherwin Vishesh Jathanna. Text may be quoted with attribution. The design and source code are not licensed for reuse.

TrueRoute analyzes historical flight data, weather patterns, and passenger sentiment to predict reliability. It assigns a score to flights based on delay risks, cancellation likelihood, and user experience, helping travelers make informed choices for smoother journeys.

Booking a flight often feels like a gamble. You pick a time and a price, then just *hope* it leaves when it should, with no real way to know which flights tend to fall apart and which quietly run smooth. **TrueRoute** was built to take the guessing out of that decision. It gathers the scattered signals that actually predict a trip going sideways, historical delays, weather, airline punctuality, route congestion, and even what past passengers had to say, and distills all of it into a single, readable number: the **TrueRoute Score**. It came together over a *48-hour build* at **Hacklytics 2025: Jurassic Age**.

### One score for a messy question

Flight reliability is genuinely hard to reason about, because it depends on so many moving parts at once. The **TrueRoute Score** is our answer to that: a *holistic reliability metric* that folds delay frequency, cancellation risk, weather exposure, an airline's on-time track record, and passenger sentiment into one figure you can glance at and trust. Instead of digging through scattered stats, a traveler sees one score and immediately knows whether a flight is a safe bet or a risky one.

### The data and the model

Underneath, a **Python and Flask** backend ingests and cleans data from several sources at once: historical flight records, weather feeds, and live sentiment streams. **Pandas** wrangles all of it into shape, and **Scikit-learn** powers the predictive core, a *supervised model* trained on thousands of flights across many routes to estimate the likelihood of delays and cancellations. The system learns the patterns that actually drive disruption, things like time of year, airport congestion, carrier reputation, and even aircraft type, and turns them into probabilities rather than guesses.

### Listening to passengers, not just the numbers

Raw statistics only tell half the story, so TrueRoute also reads the *human* side. Using **natural language processing**, it analyzes real-time sentiment pulled from public platforms and passenger feedback, surfacing patterns like recurring complaints about a route, dips in satisfaction with a carrier, or emerging disruptions people are talking about *before* any official channel confirms them. Those qualitative signals get quantified and folded straight into the TrueRoute Score, giving it a more current, human-aware read on reliability than pure historical data ever could.

### Seeing why, not just what

The **React frontend**, built with **Vite**, keeps the experience simple: enter your flight details, see its score, and compare it against alternative routes and airlines. Crucially, it does not just hand you a number, it *breaks it down* into the factors behind it, like **Weather Risk**, **Airline History**, and **Passenger Sentiment**. A traveler can see at a glance that a flight scores low because of seasonal storms or a carrier's shaky on-time record, which turns the tool into something educational rather than a black box.

### The hard parts

The toughest challenge was wrangling large, unstructured data from very different sources and *aligning it all in time* so the predictions stayed accurate, which meant careful pipeline work and caching to keep things fast without losing real-time freshness. The other hard problem was the model itself: *cancellations are rare*, and rare events are exactly the ones that are hardest to predict yet most important to get right, so a lot of effort went into balancing precision and recall so those uncommon but costly disruptions did not slip through. Built with **Divyam Kataria** and **Raajveer Khattar**, TrueRoute is still a prototype, but the vision is clear: plug reliability forecasting directly into booking platforms and airport systems, so people can plan entire itineraries around *trust* rather than luck.

## At a glance

- **Timeline:** Feb 2025
- **Type:** Hackathon Project
- **Event:** Hacklytics 2025: Jurassic Age
- **Other Contributors:** Divyam Kataria, Raajveer Khattar

## Technologies

- Python
- Flask
- React.js
- Vite
- Pandas
- Scikit-learn
- Natural Language Processing (Sentiment Analysis)
- Machine Learning (Predictive Modeling)
- Weather & Flight Data APIs
- Data Visualization (Reliability Scoring Dashboard)
- Cloud Deployment (Scalable Flask APIs)

## Categories

- Web App
- AI/ML
- Accessibility
- Travel
- Hackathon Submission

## Photos

- Demoing TrueRoute to the judges at Hacklytics 2025, with the reliability dashboard open on the laptop

## Links

- [GitHub](https://github.com/sherwinvishesh/TrueRoute)
- [Devpost](https://devpost.com/software/trueroute)