1 of 8

Timely Travel

Raaj Patel, Andrew Melbourne, Jacob Thomas, Roman Gasiorowski

2 of 8

INTRODUCTION

Timely Travel helps travelers book flights to where they need to go, when they want to go using a feed-forward neural network.

3 of 8

The Unique Value Proposition

Timely Travel is the first booking application that prioritizes flight reliability and aims to disrupt the airline industry by keeping airlines accountable to travelers.

4 of 8

Vue (TypeScript) Frontend

Diverging from the mainstream Framework of React we choice VueJS to build our application. VueJS offers better developer experience while offering the same lightweight solution.

Supporting Add-Ons:

Pinia - State Management tool for VueJS

Vue-Router - Single Page Application Routing tool.

Tailwind - utility first CSS framework

5 of 8

Machine Learning

We created a feed-forward neural network to analyze 24 data attributes to predict if a flight will be on-time, marginally delayed, or probably delayed. Using this model, and improving it using KFold-Cross Validation, we were able to improve predictions to 97.1% accuracy.

Technologies:

  • Keras
  • Numpy
  • Sklearn
  • Tensorflow

6 of 8

Go Backend

Offering, high performance, a built in http library, concurrency, and a fast compiler we chose Go as our backend Language.

Supporting Add-Ons:

GORM - Go ORM, to help manage the MySQL database.

Gin-Gonic - High performance http library allowing us to easily build an API.

7 of 8

Cloud Deployment

Timely Travel was deployed to the cloud in a HA K3s cluster.

MySQL 8.0

For our data storage we decided to chose MySQL 8 primarily due to ease of use and familiarity.

Ubuntu 22.04

We chose Ubuntu 22.04 for both our K3s Monitor nodes and Worker nodes also due to familiarity

Nginx RP

Finally we also decided to reverse proxy our Go REST-API allowing for us to reduce TLS handshake time, increase security by using SSL and to allow for a cluster Go deployment.

8 of 8

THANK YOU