Scoring Assurance Caravan

School project completed between February and May 2020, in an academic setting.
The objective was to predict the propensity profile for caravan insurance, by analysing the available client data and implementing scoring techniques.

❓ Problematic

How to identify the profiles most likely to subscribe to caravan insurance, based on socio-demographic characteristics and past behaviours?

🛠️ Solution implemented

• 📊 Exploratory data analysis of client data (profile, history, etc.)

• 📈 Selection of explanatory variables through correlation and importance analysis

• 🧠 Implementation of several classification models:

  • k-Nearest Neighbors

  • Decision Trees

  • Scoring models type propensity score

• 🔍 Performance evaluation through cross-validation, confusion matrices and classical metrics (accuracy, F1, etc.)

⚙️ Technical stack

• Languages: Python

• NLP Libraries : sklearn (TF-IDF, cosine similarity), NLTK

• UI / App: Streamlit (or simple local interface)

• Data: HR documents (PDF, Word)

• Environment: local (Jupyter, VS Code)