• Time series analysis model(ARIMA) to identify patient admission in various department and hospitals of the group to use it for strategic decision like resource planning and budgeting
• The in-patient admission planning was a pain point for the management and they had problems dealing with bed allocation across medical seasons and by departments
• On analysis we found that the annual revenue could be increased by 12-15% by better bed allocation and resource planning during different seasons.
• Further going with Time series analysis to solve this problem, we found that the series had both monthly and weekly seasonality.
• We took care of both the seasonality during the model development and piloted for one of the major hospitals.
• On implementation of model development ,we found the Mean absolute percentage error(MAPE) to be between 2.5-7% for different hospitals and departments.

• Development of rare event logistic model to predict the effect of dyes used in contrast procedures on patients resulting in renal failure.
• Medically , it is known that a dye or contrast that is inserted in the patients is known to affect the kidney of 2-3% patients.
• But, we found that the event rate was much higher(6-7%) in our case.
• So,to improve patient care, we created a rare event logistic model using R wherein we assigned probability scores for each patient and classified high risk patients.

• Statistical classification of Custom house agents basis transactions and behaviors.(K-Means clustering)
• There are CHA agents authorised by Government to facilitate clearance of import of good for importers from the government.
• These CHAs sublicense their authority to further agents for better business.
• To create policies and to differentiate different types of CHAs \,w e created clusters so that the policy makes can create policies keeping in mind the different types of agents they would have to deal with.

• Fraud detection during import of goods
• For example: An importermay declare gold as silver to evade duty on silver or may undervalue it.
• We used Decision tress to create rules that would classify transactions as fraudulent or not and assist inspectors on port in fraud detection.

Development of rare event logistic model to predict the effect of dyes used in contrast procedures on patients resulting in renal failure.

Forecast the daily footfall in various departments in a hospital basis single and multiple variables.

Automate reports on daily Drug Dosage for each hospital unit.

Fraud management project for CBEC - Ministry of Finance

Carry out analytical project with the team and develop models for prediction of LOS, Bill amount, Patient safety etc.

Identification of Fraud Patterns, Anomaly detection, Fraud mitigation rules generation and forecast revenue.

Collect historical data for Autumn Winter and Spring Summer collections and build a statistical recommendations model based on it.

Calculating Employee Engagement and create reports for different units and assist in Action Planning

Forecasting the revenue from Service Tax (Ministry of Finance) for the next financial year using ARIMA statistical model.

Collect data for different variables and identify association among products based on Price, Category, Styles and attributes.

To create rules in SAS Fraud Framework those were to be implemented to detect fraud real time.