Portfolio
Selected Projects
Identifying Environments is Conducive to Insider Trading || Client: Fidelity Investment, End date: May 2017
Project Location: Albuquerque, New Mexico, USA Team size: 5
Project Description: Fidelity is a multinational financial services corporation based in Boston. They asked to find an environment where insider trading (if someone from an organization takes advantage of having internal financial data) can occur. The algorithm they used at that time is story stock, which based on the major change in stock price.
Hypothesis: There are so many different elements that can affect the stock price. So, if we can develop a new tool that brings together financial statement analysis, corporate events, social media, and stock price, then we can get more grip to sense insider trading.
Approach: Applying exploratory data analysis, forecasting, Statistics, NLP, and visualization. Extensively used R and Tableau. Data profiling, data preparation, experimenting time series analysis.
Result: We found more than 83% accuracy in finding the environment where insider trading may occur.
Project URL: https://www.youtube.com/watch?time_continue=2&v=X2gRouOh2gk
Fraud Detection of Ecommerce Transaction || Client: Client X, End date: June 2019
Project Location: New Orleans, LA, USA Team size: 4
Project Description: Fraud is a billion-dollar business, and it is increasing every year. In this project, we wanted to detect the fraud transaction from the historical data of the financial firm.
Hypothesis: Our hypothesis was AI-based predictive analytics to detect fraud transactions by using multiple supervised machine learning algorithms that will identify fraud with better accuracy.
Approach: We build the AI Solution, which can be used as a utility to recognize eCommerce fraud activities that will help millions of people and organizations to make safe and trusted transactions thus, reducing fraud loss. We developed the whole end to end solution in Azure Cloud Platform.
Result: We used the classification ensemble ML algorithm from python and Azure Machine Learning studio in the Azure Data Factory pipeline. The achieved accuracy was 94% and Area Under Curve 91%.