All my life, I have always been a voracious learner seeking new skills and growth. I also had a longstanding interest in artificial intelligence (AI), and machine learning (ML). So when I came across a training opportunity focusing on AI, powered by FAI Institute, I had to apply. I was accepted.  

During the program, we were all assigned projects, and mine was titled "Loan Default Risk Prediction using XGBoost."

Loan Default Risk Prediction, a machine learning solution, aims to ​​curb the substantial debt risks faced by financial organizations when providing loans. Using ML and AI we could reduce the risk the banks take by predicting the likelihood of a loan default. With the help of this innovation, there is a possible predictability of a likely loan default.  

Learning XGBoost

The project involved training a machine learning model using the XGBoost Classifier. Optimal for organized tabular data, it allows application to future large datasets.  

The project goal was to build a robust machine-learning model capable of predicting, with high accuracy, the likelihood of a loan default. Meanwhile, the process involved working with a large dataset from data science platform and community Kaggle, which includes demographic data, credit history, and loan information. XGBoost has an optimized distributed gradient boosting library designed to be highly efficient and scalable.

For my part, I gathered that the trained model can be updated with new data using the Incremental Learning approach. Incremental Learning involves updating the model with new data, independent of the data previously used for training. This step is not complete without the Fairness Evaluation. A Fairness Evaluation involves assessing the model for bias across demographics. On top of having an accurate and consistent model, it was important and necessary to verify to what degree a model may use information to make biased predictions.  

Challenges faced

One of the significant hurdles I encountered was activating the Conda environment where all the implementations took place, given that I was still developing the necessary programming skills and because of the Operating System I was using. In addition, I encountered several obstacles that nearly derailed my progress.  

Some were about setting up the stock and Intel environments, others were about the need to split data sets into 4 equal batches for the Incremental Learning.  

My laptop was struggling with the hard drive requirements of the project so it would occasionally freeze, making the tasks take longer. One solution would have been using virtual space, but I had gone too far with ‘interacting’ with my dataset and I was running out of time, so I resorted to deleting files to create space.  

My mentors were extremely helpful because there were several instances where I didn’t understand why my output was not aligned with the resource I was replicating and each time, they guided me through my mistakes and the solutions I needed.

The future of AI in Africa

This exposure has instilled in me a strong hope in the potential of artificial intelligence to revolutionize the way African countries operate and interact with the world. I believe that AI can help solve some of the continent's most pressing challenges, such as healthcare, education, and infrastructure. It can also create new opportunities for businesses, entrepreneurs, and innovators.  

Africa can leverage the power of AI and ML to foster financial inclusion, creativity and mentorship. I believe this can be achieved through more technology hubs, as well as technical empowerment in local startups.  

AI can revolutionize banking and financial services in the continent. As fintech industry exponentially grows, prediction analysis can be valuable to understand how to better serve the people while also keeping the business sustainable. AI-powered chatbots and virtual assistants can provide personalized customer support and streamline banking operations, making financial services more accessible and efficient. AI algorithms can also analyze vast amounts of financial data to identify fraudulent activities and enhance cybersecurity measures.  

By embracing AI technologies and addressing the associated challenges, I believe Africa can position itself as a leader in the global AI ecosystem, fostering innovation, economic growth, and sustainable development.

A fulfilling journey

Overall, the journey during the AI training program was fulfilling, challenging, and rewarding. It allowed me to work with real-life data, implement advanced ML techniques, and build a predictive tool with practical use in the financial services industry. The project helped me understand ML algorithms, as well as learn data analysis techniques.  

I look forward to applying the skills I learned to future projects and continuing to explore the vast domain of artificial intelligence

‍