softvis 2005: saint louis, missouri, usa michael burch, stephan diehl, peter weißgerber: visual...

Post on 20-Dec-2015

216 Views

Category:

Documents

1 Downloads

Preview:

Click to see full reader

TRANSCRIPT

SOFTVIS 2005: Saint Louis, Missouri, USA

Michael Burch, Stephan Diehl, Peter Weißgerber: Visual data mining in software archives

Martin Pinzger, Harald Gall, Michael Fischer, Michele Lanza: Visualizing multiple evolution metrics

Data Mining Terminology

• Association rules: Item changed at the same time (related item)

• Sequence rules: order of these changes

• Binary Association Rules: how often 2 items changed together

• Support: Number of transaction containing the item

• Confidence: Number of Changes for pair item over single item

• Outliers: unbalance datasets or abnormal distance

Introduction

• What is visualize

- Binary association rules

- n-ary association rules

- Sequence rules

- distribution, support and confidence –histogram

• Tool EPOSee: Integrates different view

• Purpose: detect clusters, inspect rules, zoom and

filters

EPOSee InterfacePixelmap

Support Graph

3D Bar Chartfilter

Search keywordColors

Parallel Coordinates View Decision Tree

3D branch view

Rule matrix

Item list

Rule detail Support & confidence

n-ary association rules

3D bar charts

• Strong dependecies: High Support & confidence

• Use color and heights

Visualize binary association rule only

Pixelmap

File ordering: hierarchical

Stronger related

Pixelmap Example

File coupling atdifferent directorylevel

Edges: related items

Outliers: blue

Clusters: sets of items

Support Graph

Nodes: Items

Red:high

Association Rule Matrix

y-axis: Items

x-axis: Rules

Red, blue & white pixels

Support:length

Confidencecolor

Parallel Coordinates View

VisualizeSequence

Rules

Parallel CoordinatesView

Nodes Color: Support Values

Edges Color: Confidences

Cluster on samesubdirectory

Parallel Coordinates View

Green edges: high confidence

But, no edges with high confidence is coming into these 2 nodes

Pinzger, Gall, Fischer, Lanza:Visualizing multiple evolution metrics

• Objective: Communicate the evolution of metrics of source code entities and their relationships

Kiviat Diagram

M1, M2..,M6 = 6 metrics

increasing

decreasing

Metrics

Logical Coupling

Edge: Coupling relationship

A module from Mozilla

top related