HomeSweetRoam

Inspiration & Innovation

We took inspiration from several sponsor challenges. The questions we set out to answer are: How do we build a better future with intentionality? How can we use GeoTab’s telematics data sets to help individuals find their dream location? And How do we connect people with a common interest? Introducing HomeSweetRoam. A dream location finder that adapts to you.

Functionalities

This web application has two core features. The first is to recommend a new location to stay. Through a simple survey and conversation analysis, the HomeSweetRoam application utilizes user preference data to find a location that the user is likely to love. The second feature is to connect the user with others with similar preferences. The app allows 3 different ways of doing so. They can communicate through text chat, voice call, or video call.

How It Works

How do we plan on bringing HomeSweetRoam to life? This is a diagram depicting the architecture of this application. Users can interact with HomeSweetRoam through either a web browser or a mobile browser. Browsers obtain webpages from the web server. They also contact the servers and services located on GCP or Google Cloud Platform. We will go in more details in the following slides. This is a simplified diagram for easier understanding. The main parts are the end users’ browsers, a cloud, and data from GeoTab.

GeoTab has kindly provided us with plenty of data. Some include temperature and precipitation. We store these data in a database. And through machine learning, we generate a prediction model to predict different outcomes. For example we can predict the precipitation of a location using latitude, longitude, and day of the year.

Matching & Pairing

In terms of our matching functionality, we would like to pair people with similar preferences. For example if Alice and Margret both like warm places they should connect.

How does matching work? Every user is given the option to complete a survey. Some users may choose to skip the survey. To ensure we have enough data to give a recommendation, we collect keywords used by a user during their communication with others. How do we determine keywords? Well, we can look at word frequency, length, and whether a word is a slang or professional terminology. We can use ML to create a model that models the relationship between collected keyword data and user interests from the optional survey.