Łukasz Wróbel

Data-driven adaptive selection of rule quality measures for improving rule induction and filtration algorithms

This paper presents a proposal of a rule induction algorithm selecting a rule quality measure adaptively. The quality measure plays the role of an optimization criterion of the generated rules. Nine quality measures applied by the algorithm are presented and discussed in the paper. It is shown experimentally that the proposed algorithm provides us with obtaining a classifier of the best quality. During experiments, three criteria of the classifier quality were considered: overall accuracy, balanced accuracy (average accuracy of decision classes), and complexity of the classifier (understood to mean the number of induced rules). The experiments were carried out on 34 data sets coming from the UCI machine learning repository. Moreover, a proposal of four-rule filtration algorithms is presented in the paper. Their task is to limit the number of rules in the classifier. In particular, filtration influence on the classifier quality is studied.

Data-driven adaptive selection of rules quality measures for improving the rules induction algorithm

The proposition of adaptive selection of rule quality measures during rules induction is presented in the paper. In the applied algorithm the measures decide about a form of elementary conditions in a rule premise and monitor a pruning process. An influence of filtration algorithms on classification accuracy and a number of obtained rules is also presented. The analysis has been done on twenty one benchmark data sets.

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.

Regression Rule Learning for Methane Forecasting in Coal Mines

The rule-based approach to methane concentration prediction is presented in this paper. The applied solution is based on the modification called fixed of the separate-and-conquer rule induction approach. We also proposed the modification of a rule quality evaluation based on confidence intervals calculated for positive and negative examples covered by the rule. The characteristic feature of the considered methane forecasting model is that it omits the readings of the sensor being the subject of forecasting. The approach is evaluated on a real life data set acquired during a week in a coal mine. The results show the advantages of the introduced method (in terms of both the prediction accuracy and knowledge extraction) in comparison to the standard approaches typically implemented in the analytical tools.

Rule quality measure-based induction of unordered sets of regression rules

This paper presents the algorithm for induction of unordered sets of regression rules. It uses sequential covering strategy and dynamic reduction to classification approach. The main focus is put on quality measures which control the process of rule induction. We examined the effectiveness of nine quality measures. Moreover, we propose and compare three schemes of the prediction of target attribute value of examples covered by more than one rule. We also show rule filtration algorithm for the reduction of the number of generated rules. All experiments were carried out on 35 benchmark datasets.

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.

Outlier Detection and Elimination in Stream Data – An Experimental Approach

In the paper the issue of outlier detection and substitution (correction) in stream data is raised. The previous research showed that even a small number of outliers in the data influences the prediction model application quality in a significant way. In this paper we try to find a proper complex method of outliers proceeding for stream data. The procedure consists of a method of outlier detection, a statistic used for the outstanding values replacement, a historic horizon for the replacing value calculation. To find the best strategy, a wide grid of experiments were prepared. All experiments were performed on semi–artificial data: data coming from the underground coal mining environment with an artificially introduced dependent variable and randomly introduced outliers. In the paper a new approach for the local outlier correction is presented, that in several cases improved the classification quality.

Data Intensive vs Sliding Window Outlier Detection in the Stream Data — An Experimental Approach

In the paper a problem of outlier detection in the stream data is raised. The authors propose a new approach, using well known outlier detection algorithms, of outlier detection in the stream data. The method is based on the definition of a sliding window, which means a sequence of stream data observations from the past that are closest to the newly coming object. As it may be expected the outlier detection accuracy level of this model becomes worse than the accuracy of the model that uses all historical data, but from the statistical point of view the difference is not significant. In the paper several well known methods of outlier detection are used as the basis of the model.

Influence of Outliers Introduction on Predictive Models Quality

The paper presents results of the research related to influence of the level of outliers in the data (train and test data considered separately) on the quality of a model prediction in a classification task. The set of 100 semi–artificial time series was taken into consideration, which independent variables was close to real ones, observed in a underground coal mining environment and dependent variable was generated with the decision tree. For every considered method (decision trees, naive bayes, logistic regression and kNN) a reference model was built (no outliers in the data) which quality was compared with the quality of two models: Out–Out (outliers in train and test data) and Non-out–Out (outliers only in test data). 50 levels of outliers in the data were considered, from 1 % to 50 %. Statistical comparison of models was done on the basis of sign test.

Censoring weighted separate-and-conquer rule induction from survival data.

OBJECTIVES Rule induction is one of the major methods of machine learning. Rule-based models can be easily read and interpreted by humans, that makes them particularly useful in survival studies as they can help clinicians to better understand analysed data and make informed decisions about patient treatment. Although of such usefulness, there is still a little research on rule learning in survival analysis. In this paper we take a step towards rule-based analysis of survival data. METHODS We investigate so-called covering or separate-and-conquer method of rule induction in combination with a weighting scheme for handling censored observations. We also focus on rule quality measures being one of the key elements differentiating particular implementations of separate-and-conquer rule induction algorithms. We examine 15 rule quality measures guiding rule induction process and reflecting a wide range of different rule learning heuristics. RESULTS The algorithm is extensively tested on a collection of 20 real survival datasets and compared with the state-of-the-art survival trees and random survival forests algorithms. Most of the rule quality measures outperform Kaplan-Meier estimate and perform at least equally well as tree-based algorithms. CONCLUSIONS Separate-and-conquer rule induction in combination with weighting scheme is an effective technique for building rule-based models of survival data which, according to predictive accuracy, are competitive with tree-based representations.