Bell Canada was looking for ways to help the company predict customer churn, and our developers were part of the team that built the tools to accomplish this. The suite of tools included ETL pipeline to load large portions of data in chunks into an analysis suite and use it to study the customer journey to predict when a customer might be considering leaving Bell. The team used state-of-the-art machine learning technologies to extract valuable insights from time-based input sequences.

The infrastructure developed was capable of real-time analysis, allowing Bell to query our models on the fly amidst rapidly changing market conditions. The pipeline was built using Python, using Pandas and Numpy for data processing, and using Keras and Tensorflow for prediction models. Plotly and Matplotlib were used for data visualizations. The software integrated with Bell’s existing data infrastructure, and utilized GPU resources for model training.