User:Evandrojr

I'm staff member of the Southampton University and also a PhD student in the School of Mathematical Studies Interest areas

1. Data mining (especially classification and regression trees) 2. Machine learning

TreeFit data mining demonstration

Click to below to download the presentation. When you see a strange blue or white border on the right side of the slide, click on it and it will start a software demonstration (movie clip). Presentation

My research poster

Click for full size Poster

Academic projects e-Mark, a web feedback system for student grades

Written in PHP for the School of Mathematics, now supports Latex mathematics. [project page] [screenshots] [login page]

TreeFit, a data mining sofware

Written in C# for the NHS infomation authority [project page] [screenshots]

Simbuilder, a discrete event simulation software.

Looks like a Simul8 clone, but Simbuilder includes support for event parallelism and event serialisation. Written in C# (No sponsor yet, you can be the sponsor of this project) [project page] [screenshots]

Writing the Article: Scaling classification trees: Reducing the NP-complete problem for binary grouping of classification tree splits to complexity of order n^2log(n)

Evandro Leite Paul Harper

Department of Operational Research, University of Southampton

Abstract. Decision tree is an important tool of data mining. Past approaches used for classification tree were not able to deal with independent categorical variables that had more than thirty different values and assure that the result was optimal (for binary splits). This paper proposes a technique that can be applied to a regression tree with any number of categories’ values. The result of the proposed technique is equivalent as executing the full search of all possible ways of splitting, i.e. (n^2log(n)) possibilities for each categorical variable.

Address: Faculty of Mathematical Studies University of Southampton Highfield Southampton England SO17 1BJ Office: Room: 11005 Building: 54 Office: +44 (0)23 8059 3644 Mobile: +44 (0)7737474920