Houssem Eddine Azzag

• **Data Preprocessing and Cleaning**: I worked on processing complex, multi-source SNDS data, cleaning and preparing it to ensure quality. I collaborated closely with cancer specialists to perform feature engineering, identifying key variables relevant to patient mental health based on treatment paths and sequences.

*Model Development and Evaluation**: I applied a variety of machine learning models to predict mental health outcomes, including deep learning models such as **LSTM with autoencodersto capture sequential patterns in treatment, as well as traditional models likeRandom Forestand **XGBoost**. I evaluated model performance and refined it to achieve reliable and interpretable results.*Developing SNDSPOP Package**: As part of the project, we developed a package called **SNDSPOPfor simplifying the characterization of cancer patient groups. This package enables developers and researchers to analyze patient characteristics with a single line of code. SNDSPOP provides detailed insights, such as:

• Treatment Usage: Automatically calculates the usage percentage of various treatments (e.g., radiotherapy, surgery).
• Treatment Regimen Analysis: Characterizes treatment regimens (e.g., neoadjuvant, adjuvant).
• Sankey Diagram Visualization: Generates a Sankey diagram to visualize patient treatment sequences, helping users understand treatment flows and transitions based on medical records.