JMIR Publications

ABSTRACT

Background:

Over the past years, homelessness has become a significant issue around the globe. The largest social services organization in Thunder Bay, Ontario observed that a majority of people experiencing homelessness in the city were from outside of the city or province. Thus, to improve programming and resource allocation for homelessness in the city, including shelter use, it was important to investigate the trends associated with homelessness and migration.

Objective:

This research aims to address three research questions related to homelessness and migration in Thunder Bay: (1) What factors predict if a person who migrated to the city and is experiencing homelessness stays or leaves?; (2) If an individual stays, how long are they likely to stay? and (3) What factors predict their stay duration?

Methods:

We collected the required data from two sources; one through a survey conducted with people experiencing homelessness at three homeless shelters in Thunder Bay and the other from a database of a homeless information management system. In total, records of 110 migrants were used for the analysis. Two feature selection techniques were used in addressing the first and third research questions; and to predict the stay duration of homeless migrants at shelters, eight machine learning models were used. In addition, data augmentation was performed to improve the size of the dataset and a cross-validation technique was used to avoid possible model overfitting.

Results:

Factors predicting an individual's stays included: home or previous district, highest educational qualifications, recent receipt of mental health support, migrating to visit family or friends, and finding employment upon arrival. For research question 2, among the classification models developed for predicting the stay duration of migrants, the random forest and gradient boosting tree presented better results with 0.91 and 0.93 of the area under the curve values respectively. Finally, the home district, band membership, status card, previous district, and recent support for drugs and/or alcohol were recognized as the factors predicting stay duration.

Conclusions:

The application of machine learning to make predictions related to migrants’ homelessness and to investigate how various factors become determinants of them is demanded. We hope that the findings of this study will aid future policy-making and resource allocation to better serve people experiencing homelessness. However, interpretation of the identified factors in decision-making is further required.