Modern Analytics Inc. August 2015 - at Present• Developing Incentive Program for the GM Cars: The Hive/Spark are used to acquire incentive data and the SAS predictive models with PMML codes were generated in SAS Enterprise Miner. A sequence of steps was taken to optimize the statistical models and PMML codes to comply with the web application requirements. This project proofed to be very profitable for the company.
• Developing Web Traffic Prediction Models for Marketing Purposes: This is an assignment to develop, improve and maintain the statistical forecast models for the web visits. The models are refreshed on a weekly basis along with driver analysis, and the actual and forecast data were compared to help marketing folks develop better ROI strategies in the future. Many Ad-hoc analyses were also conducted.
Western Digital Inc. September 2002 - August 2015• Analyzing Customer Feedback by Using Natural Language Processing with Python: This project collects the customer feedback from major on-line merchants through HADOOP, and applies a variety of machine learning methods in NLP. It helps the marketing people to lay out the future product development strategies to address customers’ demands from the voice of customers.
• Using Time Series Models to Improve Warranty Prediction Accuracy: This project applies time series ARIMA models in SAS/Enterprise Miner to improve the forecast accuracy of warranty returns from the customer. Through this project, the new forecasting system delivers better precision on warranty reserve and planning for future customer service demands. The Excel pivot tables were also used extensively in this project.