Some projects I worked on…

---

Natural Language Processing (NLP)

Topic Modeling on customer contacts

Project description: From contact transcripts, identify recurring and event-driven causes of customer contact to quickly take action that will improve customer experience and reduce costs associated with unwanted contact volume. In order to achieve this goal, provide a dashboard to allow easy exploration of causes, events, and topics to make it easier to understand reasons and trends regarding customer contact.

---

Computer Vision

Car Damage Detector (Instance Segmentation and Image Classification)

Project description: Automaticly detect damages on a car image and produce a damage summary report. The goal is to experiment image segmentation with Mask-RCNN to see the potential and the complexity applied to a car damage detection problem.

---

Time Series

Call Center Forecasting

Project description: Forecast the number of call offering for the next 30 days and apply ErlangC algorithm on projections to calculate the optimal number of needed resources, each 30 minutes of the day, to optimize the workforce planning.

Financial & Actuarial Forecasting

Project description: Forecast monthly and quarterly financial and actuarial indicators to support strategic planning using time-series decomposition technics.

---

Survival Analysis

Customer Churn (Group insurance)

Project description: For new submissions and renewals, estimate the conditional time to event (churn) to help actuarial decisions.

Long-term disability claims duration

Project description: Estimate the duration of long-term disability claims to support actuarial decisions and to optimize claims management.

---

Association Rules

Consumption of drugs in health insurance

---

Anomaly Detection

Health claims Fraud Detection

Project description: Highlight anomalous claims and extract data rules to understand them and help improve the overall claim analysis process.

---

Classification

Optimization of life insurance requirements (Detection of undeclared smokers)

High cost drug claimants risk profile

Long-term disability risk profile

Deterioration of experience

---

Regression

Anticipated claim amount for pricing optimization

---

Other Experimental Studies / Medium publications

• Implementing GAN for regression: Generator-Discriminator-Regressor

• Mixture Density Network (MDN) for multivariate Regression (On Medium with Python codes)

• Hands-On Advanced Deep Learning Time Series Forecasting with Tensors (On Medium with Python codes)

• RdR score metric for evaluating time series forecasting models (On Medium with Python codes)

• How I beat XGBoost with Linear Regression! (On Medium with Python codes)

• Visualization trick for multivariate regression problems (On Medium with Python codes)

• Demonstrating the power of feature engineering (On Medium with Python codes)  

---

Page template forked from evanca