The Gang Reduction and Youth Development (GRYD) program is an initiative conducted by the mayor’s office of the City of Los Angeles, with the aim to curb gang violence and to promote youth development among at-risk individuals. Eligibility for program services is established using the Youth Services Eligibility Tool (YSET) questionnaire. The goal of the study is to evaluate the effectiveness of the GRYD program through a dataset that records the participants’ responses to the YSET questionnaire. Existing machine learning algorithms, such as Linear Support Vector Machine (LSVM) and Neural Network (NN), can help us accurately predict future responses and risk-indicating scores of the GRYD program participants, yet interpreting the resulting models of these algorithms remains difficult. On the other hand, through the use of dynamical models such as Dynamic Mode Decomposition (DMD) and Dynamic Mode Decomposition with Control (DMDc), we were able to not only predict individual responses and risk scores as accurately as by using machine learning models, but also interpret how responses to each question change over time in different demographic groups of participants. In addition, we were able to observe how significant each question is when determining participants’ overall risk scores. To further improve the effectiveness of program, we suggest GRYD to consider targeting services in different risk-influencing areas differently for participants based on program progress and their distinct demographic groups.
Due to the data privacy, only some codes, the final presentation and report are included.