Senior Quantitative Analyst
Current• For forecasting suspicious transactions, built compared Facebook's Prophet based model, linearly weighted moving average and seasonal decomposition ARIMA models, among which, the latter performed the best• Applied Named Entity Recognition NLP model to analyze the flow of transactions and transacting patterns in order to find the root cause of increase in transaction volumes compared to historical trends• Used Random Forest Classifier and XGBoost to understand the relationship between predictors and suspicious transactions, in order to identify important features which predicted money laundering activity • Applied NLP model trained on financial articles to identify industries, using similarity score, to find the cause of historic wires transaction volume drop compared to historical trends.• Developed automated reports with plots and metrics to track the month-over-month trends in Alerts, Cases and SARs• Applied Stratified Random Sampling technique to obtain and evaluate massive continuous and discrete variable datasets, through Distribution Analysis of representative samples• Applied K-Means clustering to group similar transactional and customer profiles for segmentation and selected best clustering method based on cluster evaluation metrics• Implemented and presented a proof of concept new method for AML rules by combining variables and applying distribution analysis to filter suspicious transactions• Conducted statistical tests (kolmogorov smirnov, t-test) to compare transaction distributions• Extracted transactions details from unstructured dataset and conducted distribution analysis to determine new Anti-money Laundering scenario-based thresholds• Analyzed historical monthly transactions to find the cause of deviation in a given month's incoming and outgoing transaction volume and amount
