Computer Prediction in Finance

Entry requirements: Basics of big data, programming and financial terminology

Credits: 4

Course: Elective

Language of the course: Russian


Valentina Guleva


Students will learn:

  • To use methods of gathering and processing large volumes of data to solve pertinent scientific problems, to process large amounts of data in finance
  • Key approaches to building predictive models for financial processes
  • Methods of formal description of predictive models in the logic of system analysis
  • Methods of scientific programming and organization of work in MatLab, Python, R
  • To make predictions based on big data analysis in financial markets and form analytical materials
  • To master methods of interpretation of prediction results, domain-specific validation and verification of models


  • Introduction to prediction theory: Concept of prediction. Types of prediction. Predictive scenario. Predictive ability. Lead time. Point of no return and anticipation. Prediction artifacts. Over-model operations.
  • Predictive models: Types of predictive models. Causal model. Surrogate models. Hybrid models. Data Driven Approach (data-based models). Examples of predictive models of financial processes.
  • Predictive scenario: Definition of predictive scenario. Extrapolation. Interval prediction. Parameter sensitivity. Variability of scenarios. Methods of generating alternative scenarios. Examples of applications in finance.
  • Data assimilation in models: Assimilation. Assimilation in precondition. Statistic correction - assimilation in prediction result. Assimilation in model parameters. Assimilation methods. Cressman scheme. VAR assimilation methods. Kalman filter. Examples of applications in finance.
  • Ensemble prediction and consensus forecast: Definition of ensemble. Ensemble in precondition. Ensemble in alternative models. Ensemble sufficiency and consistency. Ensemble growth methods. Identification of ensemble parameters on real data. Consensus forecast. Examples of applications in finance.
  • Lead time and prediction quality: Validation and verification of predictive models. Beta testing of predictive models. Prediction accuracy criteria. Definition of lead time. Statistic evaluation criteria of lead time. Examples of applications in finance.
  • Data quality control in models: Data as object of metrological analysis. Gaps and pollution in data. Statistic criteria of data quality on universal set. Methods of data quality control based on connectivity. Methods of data quality control based on proximity measure. Methods of filling gaps in data.
  • Calibration of models based on data: Definition of calibration. Calibration as stationary parameter assimilation. Choice of calibration window. Calibration objectives. Numerical optimization methods for solving calibration problems. Multi-objective calibration and Pareto fronts. Taylor diagram. Examples of applications in finance.
  • Adaptive predictive models: Self-learning of a predictive model. Model memory and oblivion window. Model memory functions. Recursive methods of recalculating model coefficients while learning. Reference model. Stochastic monitoring based on models. Examples of applications in finance.


Lectures and labs


Attendance is mandatory. Students should complete all the assignments. The final grade is based on the student performance throughout the course.