| Hello, *|NOME|* In the last month of 2024, IMDS was pleased to present to the Municipal Department of Education (MDE) of Rio de Janeiro an innovative project that puts data science at the service of public education: predictive models that estimate the probability of failure of students enrolled in Junior High School (6th to 9th grade) in the municipality's public network. The solution, based on advanced machine learning techniques, will allow MDE to target actions to support students at higher risk of repeating as of 2025. Unlike traditional approaches, which rely on the analysis of scattered data, the predictor automates the identification of students at risk of failure in an agile and accurate way. It integrates multiple sources of information – grades, attendance, school history, school profile, among others – to provide an early diagnosis, allowing the network to act well before the problem worsens. The project began with a workshop that brought together technicians from different areas of City Hall. During the meeting, these professionals shared their knowledge about the determinants of failure with the IMDS technical team. From this survey, the main variables of the model were developed, which were tested at different levels of aggregation, considering absolute and relative values. Transforming the knowledge of technicians, who deeply understand the phenomenon of failure, into variables that can be used in statistical models was one of the main challenges – and differentials – of this project. This process required rigorous data engineering and statistical validation work, allowing for refining initial hypotheses, testing the relevance of variables, and ensuring that the factors entered into the model were representative and useful for predicting the risk of failure at the end of the school year. The solution was developed from the combination of time series techniques in data engineering and high-performance machine learning algorithms, such as XGBoost, in modeling. The first tests of the Predictor showed promising and robust results. The models demonstrated high accuracy in identifying students at higher risk of failure, enabling the simulation of different intervention scenarios. After the delivery and presentation of the results to the MDE technicians, the Secretary of Education, Renan Ferreirinha, highlighted that many times failure happens because the management is unable to offer the necessary conditions for the student to learn, and that the Predictor can drive new actions: "In order for us to make the necessary adjustments, we need to identify the students who most need our support — and the sooner the better. IMDS was able to develop an extremely innovative solution, with very robust statistical models, together with our internal team of servers, who were themselves very dedicated and analytically focused so that, in the first two months of the year, we had a very strong accuracy of the students who are facing difficulties." [Watch the video here]. The next steps of this project involve developing predictive models for Elementary School (1st to 5th grade) so that the network can start actions even earlier, when learning gaps begin to appear. In addition, we will start to strengthen the use of the predictor by offering managers, principals and teachers of the network a workshop on possible uses and challenges for the use of the predictor in their routine activities.
With the Predictor, managers and educators will be able to act in advance, ensuring that support reaches those who need it the most. The solution will allow a more efficient allocation of educational resources, enhancing the impact of programs such as school reinforcement, personalized tutoring, and individualized monitoring. Avoiding one failure has relevant impacts on the trajectory of that student during their school period, as well as on their working life, and contributes decisively to social mobility. This initiative reinforces IMDS's commitment to improving the quality of education and promoting more equitable opportunities for public school students. Until the next "IMDS Letter"! Paulo Tafner CEO | 
