Rehima1/African-Credit-Scoring-Challenge

# African-Credit-Scoring-Challenge ## Project Overview This project aimed to develop a Deep learning model to predict loan defaults in African financial markets. The goal is to build a model that accurately assesses the likelihood of loan defaults for both existing and new customers. ## Data The project utilizes three datasets: - **Train.csv:** Contains historical loan data with features like customer demographics, loan details, and target variable indicating default status. - **Test.csv:** Contains loan data for which predictions need to be made. - **economic_indicators.csv:** Contains macroeconomic indicators for different countries. ## Methodology 1. **Data Loading and Preprocessing:** - Load the train, test, and economic indicators datasets using pandas. - Merge economic indicators with train and test data based on country ID. - Handle missing values using imputation techniques (e.g., mean imputation). - Convert date columns to datetime objects and extract relevant features (e.g., month, day, year). - Encode categorical features using label encoding. - Create new features based on domain knowledge (e.g., interest rate, total amount to repay, loan repayment days). 2. **Exploratory Data Analysis:** - Analyze data distributions, correlations, and patterns using visualizations (e.g., histograms, scatter plots, heatmaps). - Identify key features that influence loan default. - Understand the relationship between economic indicators and loan defaults. 3. **Feature Engineering:** - Create new features based on existing features to improve model performance. - Examples include interaction terms, polynomial features, and aggregated features. 4. **Model Selection and Training:** - Select a sequantial deep learning model. - Split the train data into training and validation sets. - Train the model using the training data and evaluate its performance on the validation set. - I used features with strong correlation to the taget (corr > 3). 5. **Prediction and Subm …