Mixture Models & Inference (MMI) Group

The MMI Group focuses on model-based clustering, classification, and statistical learning. We develop statistically principled methods for uncovering structure in complex data, with an emphasis on finite mixture models and modern inference techniques.

The group supervises postgraduate students from mathematical statistics, actuarial science, and data science, and collaborates with researchers across these disciplines from both South Africa and abroad. Based in South Africa, the group maintains active international collaborations across Europe, North America, and beyond.

Under the guidance of Prof Johan Ferreira, many students have completed honours, master's, and PhD projects in this focus area within mathematical statistics and have continued their journey to make an impact in academic or corporate roles.

Selected contributions in mixture modeling, high-dimensional classification, and robust model-based clustering:

1. 2025. "A refreshing take on the inverted Dirichlet distribution via a mode-parameterization with some statistical illustrations."

2. 2025. "A contaminated regression model for count health data."

3. 2025. "Soft computing for the posterior of a matrix t graphical network."

4. 2025. "A Power-Cardioid candidate for wind direction modelling motivated by two South African case studies".

5. 2025. "Discrete mode-mixtures of unimodal positive distributions with an application to solar energy in South Africa".

6. 2026. "Modelling bounded count environmental data using a contaminated beta binomial regression model".

Our group has collaborations with (among others):

Prof Antonio Punzo (Italy)

Prof Cristina Tortora (USA)

Dr Salvatore Tomarchio (Italy)

Dr Eden Gross (South Africa)

Dr Arno Otto (South Africa)

Prof Janet van Niekerk (South Africa)

Prof Mohammad Arashi (Iran)

Dr Tanita Botha (Australia)

Some examples of recent (completed) masters and PhD project titles:

PhD projects:

Essays on autoregressive models with nonnormal errors

Advances and considerations of Dirichlet with an emphasis on entropy

Advancements in contaminated models: from count to continuous data

Masters projects:

Skew-Laplace candidates emanating from scale mixtures for insightful computational modelling

Alternative parametric considerations for direction and distance when modeling animal movement

Views on an adaptive wavelet graph Laplacian mixture model

A contaminated generalised t model for cryptocurrency returns

Insights into the contaminated exponential model through a Bayesian framework

Opportunities for interested students:

Honours, master’s and PhD students joining our group will gain experience in both the mathematical foundations of statistics and the practical implementation of methods for complex data. Students are encouraged to work on projects that balance theory and application, often with opportunities for collaboration across disciplines.

What can you gain?

Our group is led by researchers with expertise in mathematical statistics, clustering, and machine learning. We welcome students interested in both the theoretical and applied sides of statistical research. We regularly collaborate with other researchers within South Africa and abroad! For further reading, have a look at Prof Ferreira's site.

For further details:

Information about funding and scholarships

The Wits application process

Prospective students can contact johan.ferreira@wits.ac.za with copies of their academic transcripts, a one-page motivation of their interest to work in this team in this area, and any potential budding ideas towards a possible research topic and focus.