Melike Kahya

Working on improving the 'Shop Promise' badge by building required data models, dash- boards and experiments. The badge enables merchants to build trust with customers and increase conversion.

Worked on churn prediction, Fraud Prediction projects end-to-end including target creation, feature engineering, model development, and deployment using tree-based and deep-learning methodologies.

- Clustering customers based on total and product-based profitability - Credit risk scorecard development in an agile environment - Incorporation of external data at the new predictive models - Developing binary classification / continuous prediction models upon need - Daily usage of Python and SQL

-Call Center Project: Clustered incoming calls to a call center via tf - idf technique, based on the screen codes data of each call. This enabled to analyze incoming call easily and identify contact purposes of customers -Credit Risk Scorecard: Renewed the existing risk scorecard model of a bank, which included feature extraction, target preparation and building the binary classification model -Cybersecurity: Utilized time series modeling to detect anomalies at network log data

- Loan Pricing: Derived price sensitivity curves for different customer groups from past data and built an optimization model that takes sensitivity, risk, propensity levels as input and maximizes profit by personalizing prices - Asset estimation: Developed a heuristic approach for estimating total cash asset of customers in the sector

- Provided teaching support to the faculty in Engineering Statistics, Cost Analysis, Work- Study - As a research project scholar, developed a mathematical model to minimize transportation time of humanitarian relief from supply points to disaster areas with limited vehicle and supply capacity

Facility Location Problems, Metaheuristic algorithms

Operations Research, Data Mining, Binary and Continuous Prediction Models, Energy Production Estimation