Md Imran Email & Phone Number

Hyderabad, In

Bengaluru, Karnataka, India

As Lead Data Scientist for a prominent telecommunications client, I developed an advanced churn prediction model to identify customers likely to churn in the upcoming month. The goal was to create a superior model compared to the existing manual churn prediction model and automate data ingestion, feature generation, model training, churn prediction, and prioritization report generation. This included automated evaluation of predicted prioritization versus actual churn to gauge the solution's effectiveness.We utilized diverse data elements, including historical churn data, monthly base files, utilization data, dunning, expiry details, payment and billing records, complaints, and payment reminders. Additional revenue data corresponding to each DEL was included for prioritization purposes. Rigorous Exploratory Data Analysis (EDA) ensured data completeness and usability.The structured model development process involved model design, feature generation, and evaluation of predictive power. Multiple machine learning techniques were tested, with XGBoost emerging as the most accurate. Separate models were created for normal populations and those in churn or dunning status. The final output included churn probabilities and a prioritization strategy using revenue-based cut-offs.Deployment involved a GUI-based solution hosted on the client's servers, featuring automated data ingestion, feature engineering, model training, testing, and churn prioritization report generation. A Power BI dashboard was included for monitoring and comparing various prioritization strategies. This solution significantly enhanced the client’s ability to manage churn proactively by 35-40% compared to the previous manual model, optimize revenue retention, and streamline operations.

Bengaluru, Karnataka, India

As a data scientist, I significantly contributed to the development of a payment prediction model for a leading telecommunications service client. The objective was to enhance the company's collections processes by accurately forecasting which accounts were likely to make payments and which were at risk of default. This collaborative effort led to a model implementation that significantly improved cash flow management and reduced delinquency rates.My responsibilities included analysing three years of historical billing data, payment behaviours, complaints, and data related to dunning and payment reminders. Through comprehensive Exploratory Data Analysis (EDA), we ensured data completeness and usability. The model, designed as a Binary Classification model, focused on predicting payment likelihood. We generated features from monthly snapshots, capturing payment patterns and collectively evaluating their predictive power using Weight of Evidence and Information Value.We tested various machine learning techniques, including Random Forest, Gradient Boosting, AdaBoost, XGBoost, and Logistic Regression. After rigorous evaluation, the Gradient Boosting Model (GBM) emerged as the most accurate, achieving 80%-85% accuracy. Its high performance across segments made it the preferred choice. The successful deployment of the GBM model enabled the client to manage payments proactively, enhancing cash flow and targeting collections efforts effectively. This project underscores the importance of teamwork and collaboration in leveraging data-driven solutions for financial management and collections.

Bengaluru, Karnataka, India

Passionate Data Science intern with Proficiency in Predictive modeling, data processing and Python. Eager to apply theoretical knowledge to practical applications and develop innovative machine learning solutions.

Faridabad, India

Collected data from testing on automotive HVAC components.Analyzed the data from burned-out electric motors using python libraries like pandas ,matplotlib and identified insights to drive improvements.Built machine learning models to predict target Airflow based on testing parameters.