Schneider, Bruno
E-Mail-Adresse
ORCID
Geburtsdatum
Berufsbeschreibung
Nachname
Vorname
Name
Suchergebnisse Publikationen
Task-based Visual Interactive Modeling : Decision Trees and Rule-based Classifiers
Visual analytics enables the coupling of machine learning models and humans in a tightly integrated workflow, addressing various analysis tasks. Each task poses distinct demands to analysts and decision-makers. In this survey, we focus on one canonical technique for rule-based classification, namely decision tree classifiers. We provide an overview of available visualizations for decision trees with a focus on how visualizations differ with respect to 16 tasks. Further, we investigate the types of visual designs employed, and the quality measures presented. We find that (i) interactive visual analytics systems for classifier development offer a variety of visual designs, (ii) utilization tasks are sparsely covered, (iii) beyond classifier development, node-link diagrams are omnipresent, (iv) even systems designed for machine learning experts rarely feature visual representations of quality measures other than accuracy. In conclusion, we see a potential for integrating algorithmic techniques, mathematical quality measures, and tailored interactive visualizations to enable human experts to utilize their knowledge more effectively.
Visual Integration of Meteorological and Sensor Data for Identifying Suspicious Company Behavior
We present an approach developed in course of the VAST 2017 Mini-Challenge 2. To help the ornithologist Mitch to investigate the noxious gases emitted by the four companies south of the nature preserve, we employ a combination of interactive visualizations that allow for an exploration of the data. In this paper, we present our visual-interactive approach for analyzing suspicious patterns in the data. By taking the wind data into consideration, as well, our approach allows the retrieval of patterns in the chemical releases and identify key polluters.
Uncovering the Mistford Toxic Conspiracy
To help ornithologist Mitch in understanding the poor development of the Rose-crested Blue Pipit in terms of the VAST Challenge 2017 Grand Challenge, we apply a diverse set of custom specialized tools and out-of-the-box data analysis systems to a rich data set consisting of satellite images, gas sensor measurements, movement traces and newsletter issues. Following the Visual Analytics approach, we implement a collaborative analysis loop and are able to combine data and gain insights into the current situation of the Boonsong Lekagul Nature Preserve. Finally, we come up with a hypothesis that combines suspect observations to a coherent story of illegal disposal of toxic waste involving two companies located in the reserve's vicinity.
Integrating Data and Model Space in Ensemble Learning by Visual Analytics
Ensembles of classifier models typically deliver superior performance and can outperform single classifier models given a dataset and classification task at hand. However, the gain in performance comes together with the lack of comprehensibility, posing a challenge to understand how each model affects the classification outputs and from where the errors come. We propose a tight visual integration of the data and the model space for exploring and combining classifier models. We introduce an interactive workflow that builds upon the visual integration and enables the effective exploration of classification outputs and models. The involvement of the user is key to our approach. Therefore, we elaborate on the role of the human and connect our approach to theoretical frameworks on human-centered machine learning. We showcase the usefulness of our approach and the integration of the user via binary and multiclass classification problems. Based on ensembles automatically selected by a standard ensemble selection algorithm, the user can manipulate models and alternative combinations.
Visual Integration of Data and Model Space in Ensemble Learning
Ensembles of classifier models typically deliver superior performance and can outperform single classifier models given a dataset and classification task at hand. However, the gain in performance comes together with the lack in comprehensibility, posing a challenge to understand how each model affects the classification outputs and where the errors come from. We propose a tight visual integration of the data and the model space for exploring and combining classifier models. We introduce a workflow that builds upon the visual integration and enables the effective exploration of classification outputs and models. We then present a use case in which we start with an ensemble automatically selected by a standard ensemble selection algorithm, and show how we can manipulate models and alternative combinations.
When Individual Data Points Matter : Interactively Analysing Classification Landscapes
The selection of classification models among several options with similar accuracy cannot be done through purely automated methods, and especially in scenarios in which the cost of misclassified instances is crucial, such as criminal intelligence analysis. To tackle this problem and illustrate our ideas, we developed a prototype for the visualization and comparison of classification landscapes. In our system, the same data is given to different classification models. Classification landscapes are shown in the scatter plots, together with their geographical location on a map and detailed textual description for each data record. To enhance model comparison, we implemented interactive anchor-points selection in classification landscapes. Using those anchors, the user can manipulate and reproject the model results in order to get more comparable classification landscapes. We provided a use case with crime data, for crime intelligence analysis.
DataShiftExplorer : Visualizing and Comparing Change in Multidimensional Data for Supervised Learning
In supervised learning, to ensure the model's validity, it is essential to identify dataset shifts, i.e., when the data distribution changes from the one the model encountered at the time of training. To detect such changes, a comparative analysis of the multidimensional data distributions of the training data and new, unseen datasets is required. In this paper, we span the design space of visualizations for multidimensional comparative data analytics. Based on this design space, we present DataShiftExplorer, a technique tailored to identify and analyze the change in multidimensional data distributions. Throughout examples, we show how DataShiftExplorer facilitates the identification and analysis of data changes, supporting supervised learning.
Visual Analysis of Geolocated Echo Chambers in Social Media
In news media, echo chambers refer to situations in which information is amplified or reinforced by communication and repetition. The characteristic of an echo chamber is, therefore, the absence of controversial discussions and a narrow set of opinions about a topic. We propose the use of Visual Analytics to describe spatiotemporal distributions of echo chambers using Twitter data, for specific geolocated events, such as concerts, strikes, demonstrations, etc. We analyze the echo chambers for Boston Marathon Bombing that took place on April 15, 2013. The social groups are displayed by a matrix view containing all connected components of the tweet mention graphs. To identify similar opinions, as well as, the diversity of topics in a discussion, we apply text classification and sentiment analysis. Lastly, we present initial findings based on real-world data.
Visual analytics for inspecting the evolution of a graph over time : pattern discovery in a communication network
In this paper, we present two approaches developed to visually analyze and find patterns in a communication network. The work was done for the VAST 2015 Mini-Challenge 2 (MC2), featuring a dataset with records of timestamps as well as identification of sender and receiver of text messages. Further information included the location from which a message was sent in the fictional amusement park. In the first approach, we present the data preprocessing pipeline we used for a custom visualization. In the second approach, we present how we used available data preprocessing and visualization software to get a quick and clear overview of the problem, and how we used the generated findings to feed our custom visualizations.