Brief
At Vodafone-Idea Limited, my role focused on addressing customer churn following Jio’s entry into the telecom market. I was tasked with identifying the top decile of customers who had submitted port-out requests, aiming to retain high-value users. This required working with a massive corpus of unstructured data collected from various sources and transforming it into a usable format. From over 400 features, I selected around 50 key features relevant to predicting churn.
Additionally, I worked on identifying common network issues faced by customers, particularly through social media, focusing on Twitter. My goal was to analyze sentiment, pinpoint recurring problems, and determine the regions most affected.
Approach
I started by conducting exploratory data analysis (EDA) to understand the dataset, remove outliers, and select appropriate imputation techniques for missing data. Using Gradient Boosting for classification, I successfully identified the top decile of customers at risk of churn from a dataset of 13 million Vodafone-Idea users across northeastern states.
For the social media analysis, I developed a fully automated dashboard using the Twitter API. This tool retrieved tweets for specific regions based on keywords like “Vodafone” and “Idea,” then performed topic modeling, unigram, and bigram analysis to extract key issues. The insights were visualized, highlighting problem areas and enabling corrective actions to improve network performance.
Reflection
This internship was my first experience applying Data Science and Machine Learning in the Indian telecom sector. It gave me a deeper understanding of the challenges faced by telecom companies post-Jio’s entry and how they work to retain customers. It was both insightful and exciting to contribute to such a dynamic industry.