Predicting real estate tenant occupancy

1. A system comprising:

a rental unit allocation portal configured to:

receive tenant application information associated with a tenant,

receive neighborhood data for a neighborhood in which a rental unit is located,

receive additional data including at least one of property data, spatial or property specific reasons, tenant credit information, social or online tenant activity, or prior rental history of the tenant,

train one or more machine learning models with the tenant application information, the neighborhood data, and the additional data to determine a length of stay prediction score, and

allocate a rental unit based on the length of stay prediction score; and

a prediction unit communicatively coupled to the rental unit allocation portal and configured to:

receive the tenant application information from the rental unit allocation portal;

receive the neighborhood data;

receive the additional data;

determine the length of stay prediction score based on the tenant information, the neighborhood data, and the additional data by using the one or more machine learning models, wherein each of the one or more machine learning models is selected from a group consisting of a logic regression model, a survival analysis model, a tree-based model and a gradient boosting model, wherein the rental unit allocation portal is further configured to output the length of stay prediction score to a user.

2. The system of claim 1 , wherein the prediction unit is configured to determine the length of stay prediction score by using multiple models by:

determining a respective prediction score for each of the multiple models; and

voting the prediction scores based on output from one or more of the multiple models.

3. The system of claim 2 , wherein the prediction unit is further configured to determine an accuracy of the one or more of the multiple models by using a conformal predictor to determine a predictive interval.

4. The system of claim 2 , wherein the prediction unit is configured to determine that the tenant has a likelihood of going into default during a tenancy period if each respective prediction score for each of the multiple models indicates that the tenant has a likelihood of going into default during the tenancy period or understay a predicted tenancy duration.

5. The system of claim 2 further comprising a risk allocation system configured to determine a risk for a portfolio of underlying properties with tenants or a portfolio of loans secured by tenanted properties based on the length of stay prediction score.

6. The system of claim 5 , wherein the respective prediction score for each of the multiple models is further determined based on the risk for the portfolio of underlying properties with tenants or the portfolio of loans secured by tenanted properties.

7. A method comprising:

at a rental unit allocation portal:

receiving tenant application information associated with a tenant;

receiving neighborhood data for a neighborhood in which a rental unit is located,

receiving additional data including at least one of property data, spatial or property specific reasons, tenant credit information, social or online tenant activity, or prior rental history of the tenant,

training one or more machine learning models with the tenant application information, the neighborhood data, and the additional data to determine a length of stay prediction score;

allocating a rental unit based on the length of stay prediction score;

at a prediction unit communicatively coupled to the rental unit allocation portal:

receiving the additional data; and

determining the length of stay prediction score based on the tenant information the neighborhood data, and the additional data by using the one or more machine learning models, each of the one or more machine learning models is selected from a group consisting of a logic regression model, a survival analysis model, a tree-based model and a gradient boosting model; and

outputting the length of stay prediction score to a user from the rental unit allocation portal.

8. The method of claim 7 , wherein determining the length of stay prediction score comprises:

determining a respective prediction score for each of the multiple models; and

voting the prediction scores based on output from one or more of the multiple models.

9. The method of claim 8 , wherein determining the respective prediction score further comprises determining an accuracy of corresponding model by using a conformal predictor to determine a predictive interval.

10. The method of claim 8 further comprising determining that the tenant has a likelihood of going into default during a tenancy period if each respective prediction score for each of the multiple models indicates that the tenant has a likelihood of going into default during the tenancy period or understay a predicted tenancy duration.

11. The method of claim 8 further comprising determining a risk for a portfolio of underlying properties with tenants or a portfolio of loans secured by tenanted properties based on the length of stay prediction score.

12. The method of claim 11 , wherein the respective prediction score for each of the multiple models is further determined based on the risk for the portfolio of underlying properties with tenants or the portfolio of loans secured by tenanted properties.

13. One or more computer-readable non-transitory storage media, the media encoded with instructions for performing operations comprising:

receiving tenant application information associated with a tenant;

receiving neighborhood data for a neighborhood in which a rental unit is located;

receiving additional data including at least one of property data, spatial or property specific reasons, tenant credit information, social or online tenant activity, or prior rental history of the tenant;

training one or more machine learning models with the tenant application information, the neighborhood data, and the additional data to determine a length of stay prediction score;

determining the length of stay prediction score based on the tenant information, the neighborhood data, and the additional data by using the one or more machine learning models, each of the one or more machine learning models is selected from a group consisting of a logic regression model, a survival analysis model, a tree-based model and a gradient boosting model; and

allocating a rental unit based on a length of stay prediction score; and

outputting the length of stay prediction score to a user via the rental unit allocation portal.

14. The media of claim 13 , wherein determining the length of stay prediction score comprises:

determining a respective prediction score for each of the multiple models; and

voting the prediction scores based on output from one or more of the multiple models.

15. The media of claim 14 , wherein determining the respective prediction score further comprises determining an accuracy of corresponding model by using a conformal predictor to determine a predictive interval.

16. The media of claim 14 , wherein the operations further comprise determining that the tenant has a likelihood of going into default during a tenancy period if each respective prediction score for each of the multiple models indicates that the tenant has a likelihood of going into default during the tenancy period or understay a predicted tenancy duration.

17. The media of claim 14 , wherein the operations further comprise determining a risk for a portfolio of underlying properties with tenants or a portfolio of loans secured by tenanted properties based on the length of stay prediction score.

18. The media of claim 17 , wherein the respective prediction score for each of the multiple models is further determined based on the risk for the portfolio of underlying properties with tenants or the portfolio of loans secured by tenanted properties.

19. The media of claim 13 , wherein the neighborhood data includes at least one of a desirability of the rental unit, a location of the rental unit, school data, crime rate data, or median income data.

20. The media of claim 13 , wherein determining the length of stay prediction score comprises:

determining a respective prediction score for each of the multiple models;

determining normalized prediction scores from the prediction scores for each of the multiple models;

determining an average length of stay prediction score based on the normalized prediction scores.