Marek Grzegorowski

Mining Data from Coal Mines: IJCRS’15 Data Challenge

We summarize the data mining competition associated with IJCRS’15 conference – IJCRS’15 Data Challenge: Mining Data from Coal Mines, organized at Knowledge Pit web platform. The topic of this competition was related to the problem of active safety monitoring in underground corridors. In particular, the task was to design an efficient method of predicting dangerous concentrations of methane in longwalls of a Polish coal mine. We describe the scope and motivation for the competition. We also report the course of the contest and briefly discuss a few of the most interesting solutions submitted by participants. Finally, we reveal our plans for the future research within this important subject.

Scaling of Complex Calculations over Big Data-Sets

This article introduces a novel approach to scale complex calculations in extensive IT infrastructures and presents significant case studies in SONCA and DISESOR projects. Described system is enabling parallelism of calculations by providing dynamic data sharding without necessity of direct integration with storage repositories. Presented solution doesn’t require to complete a single phase of processing before starting the next one, hence it is suitable for supporting many dependent calculations and can be used to provide scalability and robustness of whole data processing pipelines. Introduced mechanism is designed to support case of still emerging data, thereby it is suitable for data streams e.g. transformation and analysis of data collected from multiple sensors. As will be shown in this article, this approach scales well and is very attractive because can be easily applied to data processing between heterogeneous systems.

RDBMS Model for Scientific Articles Analytics

We present the relational database schema aimed at efficient storage and querying parsed scientific articles, as well as entities corresponding to researchers, institutions, scientific areas, et cetera. An important requirement in front of the proposed model is to operate with various types of entities, but with no increase of schema’s complexity. Another aspect is to store detailed information about parsed articles in order to conduct advanced analytics in combination with the domain knowledge about scientific topics, by means of standard SQL and RDBMS management. The overall goal is to enable offline, possibly incremental computation of semantic indexes supporting end users via other modules, optimized for fast search and not necessarily for fast analytics, as well as direct ad-hoc SQL access by the most advanced users.

Massively parallel feature extraction framework application in predicting dangerous seismic events

In this paper we introduce an automated mechanism for knowledge discovery from data streams. As a part of this work, we also present a new approach to the creation of classifiers ensemble based on a wide variety of models. Furthermore, we describe an innovative, highly scalable feature extraction and selection framework designed to work with the MapReduce programming model and the application of designed framework to build an ensemble of classifiers which takes into account both the quality and the diversity of individual models. The effectiveness of the solution has been verified through a participation in an open data mining competition which concerned the problem of predicting periods of increased seismic activity causing life-threatening accidents in coal mines. The submitted solution obtained the highest AUC score of all the solutions uploaded by 106 participating research teams.

Window-based feature extraction framework for multi-sensor data: A posture recognition case study

The article introduces a novel mechanism for automatic extraction of features from streams of numerical data. It was originally designed for the purpose of processing multiple streams of readings generated by sensors in coal mines. The original research was conducted on methane concentration analysis in the DISESOR project. The article demonstrates an application of the elaborated mechanism for the case of tagging short series of readings from sensors that monitor activities and movements of firefighters during the action with labels corresponding to firefighter activities. The purpose of the experiment was to assess how the automatic feature extraction and construction of classifiers (without parameters tuning and without the use of classifier ensembles) can cope with the competitions task in comparison to other participants.

Window-Based Feature Engineering for Prediction of Methane Threats in Coal Mines

We present our results of experiments concerning the methane threats prediction in coal mines obtained during IJCRS’15 Data Challenge. The data mining competition task poses the problem of active monitoring and early threats detection which is essential to prevent spontaneous gas explosions. This issue is very important for the safety of people and equipment as well as minimization of production losses. The discussed research was conducted also to verify the effectiveness of the feature engineering framework developed in the DISESOR project. The utilized framework is based on a sliding window approach and is designed to handle numerous streams of sensor readings.

Predicting seismic events in coal mines based on underground sensor measurements

Abstract In this paper, we address the problem of safety monitoring in underground coal mines. In particular, we investigate and compare practical methods for the assessment of seismic hazards using analytical models constructed based on sensory data and domain knowledge. For our case study, we use a rich data set collected during a period of over five years from several active Polish coal mines. We focus on comparing the prediction quality between expert methods which serve as a standard in the coal mining industry and state-of-the-art machine learning methods for mining high-dimensional time series data. We describe an international data mining challenge organized to facilitate our study. We also demonstrate a technique which we employed to construct an ensemble of regression models able to outperform other approaches used by participants of the challenge. Finally, we explain how we utilized the data obtained during the competition for the purpose of research on the cold start problem in deploying decision support systems at new mining sites.

SONCA: Scalable Semantic Processing of Rapidly Growing Document Stores

Scientific data constitutes a great asset. However, its volume is far bigger than any human can comprehend. Therefore, automatic analytical, search and indexing solutions are called for. In this paper we present the architecture and the data model of such a system. SONCA (Search based on ONtologies and Compound Analytics) is a platform to implement and exploit intelligent algorithms identifying relations between various types of objects (publications, inventions, scientists and institutions). The results of these algorithms can be used to build semantic search engines but also can be fed into further analytical algorithms in order to find even more associations.We also show experimental evaluation of the performance of SONCA. Its results are promising and we argue that SONCA’s architecture is robust.

Governance of the Redundancy in the Feature Selection Based on Rough Sets’ Reducts

In this paper we introduced a novel approach to feature selection based on the theory of rough sets. We defined the concept of redundant reducts, whereby data analysts can limit the size of data and control the level of redundancy in generated subsets of attributes while maintaining the discernibility of all objects even in the case of partial data loss. What more, in the article we provide the analysis of the computational complexity and the proof of NP-hardness of the n-redundant super-reduct problem.

A framework for learning and embedding multi-sensor forecasting models into a decision support system: A case study of methane concentration in coal mines

Abstract We introduce a new approach for learning forecasting models over large multi-sensor data sets, including the steps of sliding-window-based feature extraction and rough-set-inspired feature subset ensemble selection. We show how to integrate this approach with the major data-processing-related components of DISESOR – a decision support system which is a coherent and complete framework for exploring streams of sensor readings registered in underground coal mines. As a case study, we report our experiments related to the task of methane concentration forecasting. The contributions in this paper refer to both the analysis how the nature of sensor readings influenced the architecture of the developed system and the empirical proof that the designed methods for data processing and analytics turned out to be efficient in practice.