AppraisEstate

Inspiration

Racial discrimination in appraisal home values is something that has been growing over several decades The traditional way to appraise houses is appraisers walking through the properties, observing the property, collecting data and using questionable pricing models This method not only leaves room for human biases but is incredibly costly as well. In addition to this, I've used Swift a lot (UIKit mostly) but wanted to try out Apple's newer development environment called SwiftUI that lets you programmatically create all views.

What it does

We propose an application that is capable of minimising human involvement and biases in the appraisal process and boosts the accuracy of the real estate market price prediction models We seek to extract the patterns between the inputs (such as the number of bedrooms in the house) and output (price of the houses). Tried different machine learning algorithms: random forest, linear regression, K nearest neighbour on different parts of the United States. It also takes a region as input (note: 1= Northeast, 2=Midwest, 3=South, 4=West, 5=not sure).

How we built it

Swift and objective C are used to design the front end of the app. We use TensorFlow as the machine learning library to tackle the linear regression modelling and then use that through Apple's CoreML API interface.

Challenges we ran into

Data collection was something that we struggled on quite a bit. Data on new housing is ready available by the US Department of Housing, however data on older houses is harder to get and we had to web scrape appraisals to collect that data. Some of the calculations are currently dubious as the data source I have for now is incredibly limited.

Accomplishments that we're proud of

We are proud of having a working prototype that demonstrates the potential of this technology in removing housing biases in the future.

What we learned

We learnt the value of data collection, cleaning and being able to solve a smaller subset of problems within a large issue such as the housing market.

What's next for AppraisEstate

I plan to continue developing this application, collecting more data, and eventually making this into a market place (initially aimed at the US and then expand it to other regions in the world)

Built With

• ai
• api
• coreml
• objective-c
• swift
• tensorflow