Predicting Rent with Machine Learning: Smarter Choices for Tenants and Landlords

(This was a part of the Research Adventure Program at Research Spark Hub Inc.,

As the population grows and prices go up, it is crucial for people to manage their finances, especially when it comes to their homes. Many different types of people rent apartments or houses, including young university students, families whose kids need a good education, seniors who live alone or in assisted living, etc. What can help all renters however is predicting rent for their homes. Rent prediction is useful to both tenants and landlords. Tenants can make more informed decisions about where they choose to live, and landlords can set fair prices to regulate the housing market.

Could machine learning help to predict the rent?

Many things can impact a place’s rent, such as its features or surroundings. These can include the location, size, distance from local transport, places of work, amenities, stores, and restaurants. Ranking features by their importance is very helpful to machine learning because it helps it see patterns to make more accurate predictions. Some things, such as size, are much more important to people meaning that rent prices rely a lot more on size than less important features. Size is usually one of the first features people take into consideration when looking to rent somewhere. The size of a house also allows more of other features, such as more bedrooms and bathrooms, which are also ranked high in importance, and having things like laundry available in the apartment. Another major feature people consider is the location. Rent in cities is usually higher than rent in more suburban areas, while also having smaller sizes and fewer features, but the number of nearby stores and amenities is much higher. Renters consider these features when choosing where to live, and they are especially crucial when predicting rent.

Machine learning can take this data and use it to notice trends elsewhere. Some things may not be important to most people but are crucial to a specific group. A good example of this is people with disabilities. Many disabled people have to use wheelchairs to get around, and if their homes don’t have wheelchair access, they can’t enter their homes easily. The table below shows the ranking of features for the Kaggle dataset:

Use of Machine Learning Model:

The model can be trained on available data of real-world apartments or homes and their features after the importance of each feature is ranked. Once the model analyzes the data, it can be tested to see if it can predict the rent of homes using only their features. Each time the model is tested, it gets more and more accurate until a certain number of tests, also known as epochs, are completed, which means that even if the model is tested more times, its accuracy will either barely increase, stay the same, and in some cases even lower accuracy a little bit. The accuracy is measured by the difference between the model’s output (prediction), and the actual rent of the test data. The lower the difference, the more accurate the model is. Accuracy can also be changed in other ways, like increasing the amount of layers the model has for more accuracy. Layers are what take the inputs, such as the apartment features, and process and analyze them to get the output. More layers mean that data is processed better, leading to more accurate predictions. There are also ways that accuracy may be lowered. One way is if the rate at which epochs are being completed is too high. Accuracy is lowered because it doesn’t have enough time to process the data well. Another way is if there are too many epochs. This can lower accuracy because if the model reaches a point where accuracy stops increasing, and keeps testing, it can cause the model to learn and recognize data worse. And after testing the model, it was able to get down to a loss of only 0.849, with prediction accuracy within approximately 84 USD, making the model exceptionally accurate.

To summarize, predicting rent using machine learning is a great tool that benefits both renters and landlords by helping them make informed, fair, and financially sound decisions. By analyzing important features like size, location, and accessibility, models can learn patterns that affect rental prices and provide accurate predictions. As the population grows and housing becomes more expensive, this technology can be useful to many in managing housing costs and ensuring people find homes that meet their needs.