AI Research Group
A multidisciplinary research and development unit that aims to analyze, interpret, and transform large-volume, diverse, and rapidly flowing data generated in the healthcare field into meaningful artificial intelligence models
We adopt a systematic approach to understand the complex structure of health data. We develop clinical and administrative decision support systems by integrating data from electronic health records, population-based health statistics, biobanks, and national health systems.
While ensuring the systematic implementation of data-driven health solutions, we prioritize data security, privacy, and ethical use principles. The application of big data technologies in healthcare requires special approaches, particularly in terms of scalability and performance.
The group is developing algorithms that detect inconsistencies in electronic health records, meaningfully complete missing data, and ensure data integrity. This process also includes standardization and harmonization of data from different healthcare institutions.
Core application areas we develop in health data science and big data analytics
EHR analysis, clinical pattern extraction, and patient profiling
Population-based health data modeling and disease burden analysis
Cost optimization and health expenditure analysis
Hospital capacity and healthcare personnel optimization
Seasonal disease trend and prevalence prediction models
Anonymization and data security algorithms
Regional health status mapping and spatial analysis
Drug safety monitoring and supply chain optimization
Our big data analytics systems used and tested at the national level
Predicts diabetes prevalence across 81 provinces with 94% accuracy using E-Nabız data and supports health planning with 5-year projections.
Analyzes Medula data, seasonal trends, and demographic factors to predict hospital admissions on a weekly basis with 89% accuracy.
Analyzes national e-prescription data to monitor antibiotic usage patterns and regionally predicts resistance development risk.
Analyzes symptom-based admission data in real-time to regionally assess epidemic risks and provide early warnings to health authorities.
Integrates patient admission data, treatment costs, and demographic factors to analyze the main dynamics of healthcare expenditures.
Analyzes national cancer screening data to identify risk groups, evaluates screening effectiveness, and develops personalized screening recommendations.
The methodology we use in the processes from collection to analysis of health data
Data collection from national systems such as E-Nabız, Medula, and e-Prescription is a sensitive process. We use robust integration algorithms in the face of different formats, coding systems, and data quality issues.
We use differential privacy and homomorphic encryption techniques to protect patient privacy. We focus on conducting scientifically valuable analyses while protecting personal information.
Missing values, outliers, and systematic errors are critical issues in big data analysis. We improve data quality with machine learning-assisted data cleaning algorithms.
We reveal complex relationships in health data through time series analysis, survival analysis, and multivariate statistical methods. Bayesian approaches provide uncertainty quantification.
We develop prediction models from health data using deep learning, ensemble methods, and transfer learning. Overfitting and bias problems are addressed with specialized validation strategies.
We transform complex analysis results into comprehensible dashboards and interactive visualizations for decision-makers. We provide continuous monitoring with real-time monitoring systems.
Expert fields we collaborate with through a multidisciplinary approach in health data science
Core research topics we focus on with a comprehensive and systematic approach in health data science
Health data has high-dimensional, sparse, heterogeneous, and time-dependent characteristics. We develop algorithms that detect missing values, measurement errors, and systematic biases.
Seasonal variations in disease prevalence, epidemic waves, and long-term trends are modeled with temporal analysis techniques. We develop hybrid prediction models.
We research differential privacy, homomorphic encryption, and secure multi-party computation techniques that enable scientific research while protecting patient identities.
We provide comprehensive analysis capabilities by combining structured, semi-structured, and unstructured health data. Laboratory, clinical notes, and imaging data are integrated.
We provide effective solutions for rare diseases and small patient groups by adapting models trained on large datasets to small and specific datasets.
We support national and international data sharing by developing data models and transformation algorithms compatible with international standards like FHIR, HL7, and ICD-10.
Concrete achievements in health data science and the impact we create at the national level
Would you like to collaborate with us to develop big data analytics solutions in health data science?
Strategic collaborations with Ministry of Health and national data systems
Inter-university big data research projects and doctoral programs
Solution development with cloud computing and big data technology companies