You can use Actable AI's no-code Data Science platform to predict the rental or sale price of real estate properties. A large real estate agency has a dataset of current properties and properties they rented in the past. The agency team would like to understand what are the most important factors that affect final rental prices and wonder if they can predict current rental prices of new properties so they can assure they won't rent much cheaper than the optimal market rents.
After uploading the dataset into Actable AI, the agency team can use correlation analysis and quickly find that the number of rooms, area, number of bathrooms are positively correlated to final rental prices and days on the market are negatively correlated. Perhaps it's not surprising but more re-assuring of what an experienced real estate agent already knows. This insights will allow an optimal pricing so the property rents out faster at desirable market value as staying longer on the market would have negative consequences for the price and agency's revenue.
The correlation analysis also shows linear regression for each correlated factor. With a quick glimpse at the regression coefficient, the analyst knows that on average, an increase of one bedroom increases the rental prices of $1,079 per month. An extra 1 sqft increases the rental price by $3.3 per month and each day on the market drops the rental prices by $21 per month. While correlation analysis is helpful in showing linear relationships, it doesn't reveal all relationships of predicted target and its predictors. In order to reveal more complicated relationships, one can use our non-linear regression analysis to predict rental prices and predictability of variables.
A regression analysis models and predicts a continuous variable (e.g. salary, price, sales). Using Actable AI's regression analysis, the user can simply choose rental price as a prediction target and other properties as predictors, anyone can easily build a prediction model and generate the predictions with a few clicks. Once the analysis finishes running, it will return predicted results and how much each predictor contributes to the final prediction (Shapley values). For example, the highlighted row shows the predicted rental price of the property is $4,734.6. The fact that it has 3 bedrooms adds $873.4 compared to the average rental price and the location in downtown reduces its price by $1.82.
The analysis also shows its performance on a validation dataset (randomly choosing from the initial data, these are not used for training the model). For example, this analysis shows a very strong predicting capability of the model R2 is 0.99 (max is 1) and the root mean square error (RMSE) is about $45.3. Biggest predictors for rental prices are number of rooms, number of bathrooms and locations. The importance ranking shows how important is each variable for the predictions.
The dataset consist of example rental properties with 7 variables and over 5000 rows. It can be used to run various predictions and missing value completions. The dataset below is full and you can delete some parts for your prediction project. Below are the 7 variables.
number_of_rooms: total number of rooms in the property
number_of_bathrooms: total number of bathrooms
sqft: total area of the property
location: location classified as "great", "good", and "poor"
days_on_market: days from listing to rental
neighborhood: the actual neighborhood
rental_price: an actual rental price
