Machine Learning in Finance II 2400-QFU2MLF2
The course consists of 3 chapters divided according to the class of presented algorithms: 1) boosting models 2) (deep) neural network models 3) bayesian time series models. It is conducted in the form of interactive laboratories with the use of case studies which are carried out in parallel with the lecture part.
Chapter 1. Boosting models
1. AdaBoost
2. Gradient Boosting
3. eXtreme Gradient Boosting
4. Light Gradient Boosting Machine
5. CatBoost
Case study - cross sell banking/insurance product - propensity to buy models
Chapter 2. Neural network models
1. Multilayer Perceptrons
2. Recurrent Neural Network
3. Convolution Neural Network*
4. Attention mechanism in Neural Network
Case study 1 - forecasting the demand for products in large-format stores
Case study 2* - car damage classification in the insurance company
Chapter 3. Bayesian time series models
1. Facebook Prophet
2. Uber Orbit (as a framework)
Case study - forecasting the volume of parcels delivered by a logistics company
Chapter 4. Ensembling methods*
Project presentation
Type of course
Course coordinators
Learning outcomes
After completing the course, the students will have structured and reliable knowledge on boosting models, neural networks, and Bayesian time series models. They will be able to apply them for both regression and classification problems. They will know the theoretical foundations of these algorithms, as well as have programming skills allowing them to deploy the models in practice, also in the cloud framework. They will also know how to interpret results and explain how they work to other non-technical people.
K_W01, K_U01, K_U02, K_U03, K_U04, K_U05, KS_01.
Assessment criteria
Preparing two machine learning projects were prepared in groups of at most 2 students - one for regression problem and one for classification. Each project should be prepared on a different dataset selected by the students - one reasonably small dataset and one large dataset - approved by the tutor (for example from https://www.kaggle.com). Students are to prepare a presentation and an extended report in a Python notebook, containing blocks of code that will allow the teacher to fully reproduce the applied analysis.
The following weights are used to determine the final grade:
20% - Presentation
80% - Extended report
The threshold to pass is equal to 60%.
Additional information
Additional information (registration calendar, class conductors, localization and schedules of classes), might be available in the USOSweb system: