Representation of Complex Data Types for Machine Learning
|Title||Representation of Complex Data Types for Machine Learning|
|Summary||Finding ways to represent complex data types (e.g. histograms) present in Logged Vehicle Database databse for machine learning-based fault prediction|
|Keywords||Data Mining, Knowledge Representation|
|References|| Statistical Relational Learning
|Prerequisites||Cooperating Intelligent Systems and Learning Systems courses|
|Supervisor||Sepideh Pashami, Sławomir Nowaczyk|
In the ReDi2Service project we are analysing data from Volvo's "Logged Vehicle Data" Database, which contains on-board signals collected during workshop visits from ~75.000 Volvo trucks.
The goal is to use this information to predict faults and find interesting and atypical usage patterns.
However, some of the data available is in the structured form, most commonly that of histograms. The thesis project is about investigating ways of representing those histograms in ways that would make them suitable input for machine learning algorithms.
The first step is to investigate domain-independent representations, but the most interesting aspect of the project will be to make use of the available data and try to learn best representations taking into account similarities and differences among different vehicles.
More details to come...