PROJECT OUTPUT
Intelligent Railway Chatbot
Built a conversational railway assistant that can search for ticket options, support delay prediction, and retrieve expert disruption guidance for railway scenarios.
WHAT I BUILT
A railway chatbot with search, prediction, and expert guidance.
The system was designed as a practical railway assistant. It collects missing details through conversation, searches for ticket information, predicts delayed arrivals, and returns focused disruption advice from stored operational knowledge.
- Built a FastAPI railway chatbot with a browser chat interface for guided railway assistance.
- Used slot-filling conversation logic to collect origin, destination, date, time, ticket type, railcard, and return journey details.
- Integrated National Rail journey data where available, with fallback fare logic so the chatbot still returns a useful ticket result.
- Added station validation, railcard support, and return-ticket handling to make the ticket search more realistic.
- Trained delay-prediction models on historical South Western Railway data to estimate delayed train arrival times.
- Added an expert-system workflow that retrieves focused disruption guidance using structured rules, SQLite, and ChromaDB.
CAPABILITIES
Three connected AI features inside one chatbot.
Each part of the project solves a different railway assistance problem, from passenger ticket search to staff-focused disruption guidance.
| Capability | What It Does | Technical Approach |
|---|---|---|
| Cheapest ticket search | Guides users through a ticket-search conversation and returns a ticket option with a booking link. | Slot filling, station validation, National Rail API data, railcard discounts, return-ticket handling, and fallback fare logic. |
| Delay prediction | Estimates arrival time for delayed trains on the South Western Railway route. | Historical train performance data, preprocessing, model comparison, and tuned regression models. |
| Disruption expert systemExpert system | Returns focused operational advice for line blockages and station disruption scenarios. | Guided incident collection, structured contingency records in SQLite, and semantic retrieval support with ChromaDB. |
OUTPUT
Ticket search and expert-system guidance in action.
These outputs show the chatbot returning a cheapest return ticket result and retrieving focused disruption guidance for railway staff.
INTERPRETATION
The chatbot combines API search, ML prediction, and expert retrieval.
- The chatbot combines conversational search with practical railway-specific validation and fallback handling.
- The ticket workflow supports realistic user needs such as railcards, return journeys, station validation, and booking links.
- The delay-prediction task adds machine-learning support for estimating disrupted train arrivals.
- The expert-system task makes the chatbot explainable by retrieving stored operational guidance instead of inventing railway rules.