About ETRO  |  News  |  Events  |  Vacancies  |  Contact  
Home Research Education Industry Publications About ETRO

Master theses

Current and past ideas and concepts for Master Theses.

Sparse models for financial data analysis


Efficient utilization of massive financial data to support companies and individuals in strategic planning and investment decision-making is a problem of significant interest. Analyzing financial data to reveal valuable information and identify complicated patterns is a critical task in financial data analytics. Towards this goal, the extraction of features that can aid the uncovering of interesting patterns is of great importance.

In the past decade, sparse and redundant representation modelling achieved a tremendous progress in the construction and use of new signal models. Such models impose a dimensionality reduction and have gained a substantial popularity as they have been proven very efficient in many signal processing tasks.

Representing high-dimensional financial data in low-dimensional spaces has been reported to show promising results in financial applications. Existing approaches are based on the design of learned sparsifying dictionaries, i.e., overcomplete bases that enable the approximation of a high dimensional signal by a few nonzero coefficients (features). Recently, deep neural network models have become very popular as they have the ability to learn useful features of real world data, providing efficient and compact data representations (e.g. autoencoders).

Kind of work

In this thesis, we explore the expressiveness of deep neural networks to design sparse representations of financial data. The student needs to investigate existing deep learning sparse coding approaches, and propose a neural network design to produce sparse features that capture the meaningful information of volatile financial data, enabling the extraction of patterns that are useful in modern financial data analytics.

Framework of the Thesis

Possible applications in finance include stock market prediction, portfolio management, foreign exchange market prediction, and fraud detection. Depending on the application, publicly available datasets may be employed to learn appropriate sparse representations. Collecting data from open data platforms might be another option.

Number of Students


Expected Student Profile

- Computer science, applied computer science, engineering
- Good knowledge in machine learning and mathematics
- Good programming skills (Python, Matlab)


Prof. Dr. Ir. Nikolaos Deligiannis

+32 (0)2 629 1683

more info


Dr. Evangelia Tsiligianni

+32 (0)2 629 1685

more info

- Contact person




- Contact person

- Thesis proposals

- ETRO Courses

- Contact person

- Spin-offs

- Know How

- Journals

- Conferences

- Books

- Vacancies

- News

- Events

- Press


ETRO Department

Tel: +32 2 629 29 30

©2019 • Vrije Universiteit Brussel • ETRO Dept. • Pleinlaan 2 • 1050 Brussels • Tel: +32 2 629 2930 (secretariat) • Fax: +32 2 629 2883 • WebmasterDisclaimer