Mining Models for Failing Behaviors

Expert Systems with Applications, 2019

This paper proposes an effective hybrid-based methodology, called interpretable logic tree analysis (ILTA), which characterizes and quantifies event causality occurring in engineering systems with the minimum involvement of human experts. It integrates two concepts: knowledge discovery in database (KDD) and fault tree analysis (FTA). The KDD extracts the root-causes in the form of a set of interpretable (meaningful) patterns and then is exploited to automatically construct a logic tree. Only the feasible solutions consisting of non-redundant patterns that cover the maximum number of observations in the dataset are selected using a burn-and-build algorithm. These solutions are employed first to visualize the discovered knowledge under the interpretable logic tree and second, to estimate the probability of an event given the occurrence of its root-causes. An actuator system dataset is used to illustrate and validate the proposed methodology. Moreover, the ILTA methodology allows the tuning of the system states based on Bayesian control rules that characterize the nature of the discovered root-causes.

IFAC Proceedings Volumes, 1996

Fault diagnosis by means of fault trees is of considerable interest for industrial applications. The drawbacks of this approach are mostly related to the knowledge elicitation through laborious enumeration of the tree structure and ad hoc threshold selection for symptoms definition. These problems can be alleviated if a more profound process knowledge is put into play. Main idea of the paper consists of mode ling the nominal and faulty states of the plant by combining interval-like component models with first principles laws, like e.g. conservation law. Such a model serves to simulate the entire system at different faulty conditions in order to get the representative patterns of process measurable quantities in terms of training examples. To match these patterns by diagnostic rules, multi strategy machine learning is applied. As result, binary decision trees that relate symptoms to faults are obtained along with the thresholds defining the symptoms. This technique is applied to a simulated continuous process operating in the steady state and is shown to be suitable for handling incipient faults.

International Journal of Computer Applications, 2012

Defect prediction is the process of predicting the fault prone module using some pre-mined patterns or rules. Several statistical and mathematical strategies have been used in recent past to mine these rules. However, the interpretability of these rules is the matter of concern. In real development process an expert is required to demonstrate the working of mined patterns which prevents the use of these mined patterns in software development process. Considering these facts, in this study we tried to find the combination of attribute-value pair which indicates the bug. These attribute-value pair is known as defect pattern. For defect pattern mining we used GUHA (General Unary Hypothesis Automaton) procedure which is oldest yet very powerful method of pattern mining. The basic idea of GUHA procedure is to mine the entire possible and interesting hypothesis supported by the data in predefined logical form. The experimental results show that the mined patterns can be used as a rule to identify the defective module. Moreover, the mined patterns do not suffer from the interpretability problems.

Journal of Experimental Psychology: Human Perception and Performance, 1978

Fault trees represent problem situations by organizing "things that could go wrong" into functional categories. Such trees are essential devices for analyzing and evaluating the fallibility of complex systems. They follow many different formats, sometimes by design, other times inadvertently. The present study examined the effects of varying three aspects of fault tree structure on the evaluation of a fault tree for the event "a car fails to start." The fault trees studied had four to eight branches, including "battery charge insufficient," "fuel system defective," and "all other problems." Major results were as follows: (a) People were quite insensitive to what had been left out of a fault tree, (b) increasing the amount of detail for the tree as a whole or just for some of its branches produced small effects on perceptions, and (c) the perceived importance of a particular branch was increased by presenting it in pieces (i.e., as two separate component branches). Insensitivity to omissions was found with both college student subjects and experienced garage mechanics. Aside from their relevance for the study of problem solving, such results may have important implications for (a) how best to inform the public about technological risks and to involve it in policy decisions and (b) how experts should perform fault tree analyses of the risks from technological systems.

2005

Our paper presents a novel approach for identifying the key infrastructural factors determining the behavior of systems in the presence of faults by the application of intelligent data processing methods on data sets obtained from dependability benchmarking experiments. Our approach does not rely on a-priori assumptions or human intuition about the dominant aspects enabling this way the investigation of highly complex COTS-based systems. The proposed approach is demonstrated using a commercial data mining tool from IBM on the data obtained from experiments conducted using the DBench-OLTP dependability benchmark. Results obtained with the proposed technique identified important key factors impacting performance and dependability that could not have been revealed by the dependability benchmark measures.

Software fault prediction based on mining of code and design metrics has been considered by many researchers. Fault detection systems predict faults by using software metrics and data mining techniques. Various classifiers have already been used in this case; however Naïve Bayes classifier is the most commonly used. According to the results of a study performed by Lessman, no significant performance difference could be detected among the top 17 classifiers. In this paper, we will extend that study by examining the performance of 37 different classifiers in fault detection systems. We will review the results and aim to choose an appropriate classifier (Bagging) which depicts a higher performance and accuracy compared to the others. Finally, we propose a fault detection system with higher performance which manages to decrease the cost of software fault detection simultaneously. We investigate our classifier selection by evaluating the methods on a number of other datasets. Our results indicate that Bagging classifier has the highest performance in fault detection.

2001

Abstract In this paper we develop new techniques for predicting failures and monitoring in categorical event sequences. New techniques are needed because failures are rare and previous data mining algorithms were overwhelmed by the staggering number of very frequent, but entirely unpredictive patterns that exist in such databases. This paper combines several techniques for pruning out unpredictive and redundant patterns, which reduce the size of the returned rule set by more than three orders of magnitude.

2010

In statistics and data mining communities, there have been many measures proposed to gauge the strength of association between two variables of interest, such as odds ratio, confidence, Yule-Y, Yule-Q, Kappa, and gini index. These association measures have been used in various domains, for example, to evaluate whether a particular medical practice is associated positively to a cure of a disease or whether a particular marketing strategy is associated positively to an increase in revenue, etc. This paper models the problem of locating faults as association between the execution or non-execution of particular program elements with failures. There have been special measures, termed as suspiciousness measures, proposed for the task. Two state-of-the-art measures are Tarantula and Ochiai, which are different from many other statistical measures. To the best of our knowledge, there is no study that comprehensively investigates the effectiveness of various association measures in localizing faults. This paper fills in the gap by evaluating 20 well-known association measures and compares their effectiveness in fault localization tasks with Tarantula and Ochiai. Evaluation on the Siemens programs show that a number of association measures perform statistically comparable as Tarantula and Ochiai.

ACM SIGOPS Operating Systems Review, 2006

Computer problem diagnosis remains a serious challenge to users and support professionals. Traditional troubleshooting methods relying heavily on human intervention make the process inefficient and the results inaccurate even for solved problems, which contribute significantly to user's dissatisfaction. We propose to use system behavior information such as system event traces to build correlations with solved problems, instead of using only vague text descriptions as in existing practices. The goal is to enable automatic identification of the root cause of a problem if it is a known one, which would further lead to its resolution. By applying statistical learning techniques to classifying system call sequences, we show our approach can achieve considerable accuracy of root cause recognition by studying four case examples.

2013

Engine assembly is a complex and heavily automated distributed-control process, with large amounts of faults data logged everyday. We describe an application of temporal data mining for analyzing fault logs in an engine assembly plant. Frequent episode discovery framework is a model-free method that can be used to deduce (temporal) correlations among events from the logs in an efficient manner. In addition to being theoretically elegant and computationally efficient, frequent episodes are also easy to interpret in the form actionable recommendations. Incorporation of domain-specific information is critical to successful application of the method for analyzing fault logs in the manufacturing domain. We show how domain-specific knowledge can be incorporated using heuristic rules that act as pre-filters and post-filters to frequent episode discovery. The system described here is currently being used in one of the engine assembly plants of General Motors and is planned for adaptation in other plants. To the best of our knowledge, this paper presents the first real, large-scale application of temporal data mining in the manufacturing domain. We believe that the ideas presented in this paper can help practitioners engineer tools for analysis in other similar or related application domains as well.