Purpose- There is an important effort of oil and gas industry to reduce the number of accidents and incidents. There are standards to identify and record workplace accidents and incidents to provide guiding means on prevention efforts, indicating specific failures or reference, means of correction of conditions or circumstances that culminated in accident. Today, with the advances of new technologies, accidents, incidents and occupational deviations records are stored in heterogeneous repositories. Similarly, the amount of information of OHS that is daily generated has become increasingly large. Furthermore, most of this information is stored as unstructured or poorly structured data. This poses a challenge of managing big data but this is nowadays a top priority for industries. In this sense, data mining can be applied to any domain where large databases are saved to find correlation between events: accidents, incidents and occupational deviations. Some practical applications combining these techniques are: failure prediction , biomedical applications , process and quality control . Association rule learning is a popular and well-researched set of methods for discovering interesting relations between entities in large databases. It is intended to identify strong rules discovered in databases using different measures of interestingness. Many algorithms for generating association rules were presented over time. Some well-known algorithms are Apriori , Eclat  and FP-Growth , but they only do half the job, since they are algorithms for mining frequent item sets. Another step needs to be done after to generate rules from frequent item sets found in a database. In most of cases users has two main challenges during this process: i) to explore the measures of interestingness (confidence, lift, support, etc.) and ii) to understand and analyze the large number of association rules. In this sense, an intuitive visualization of mined rules becomes a key component in a decision-making process. In this paper, we propose a novel visualization of spatio-temporal rules that provides the big picture about risk analysis in a real world environment. Our main contribution lies in an interactive visualization for accident interpretation by means of well-defined spatio-temporal constraints in the offshore oil industry domain.
Design/methodology/approach- Association rules are not always simple to understand because results set are often very large and/or because the rule itself demands explanation. For the former problem, many techniques have been proposed to filter the most relevant set of rules. However, the later problem has received less attention. Many techniques are quite difficult to understand and to correlate items making hard for the user to take a decision.
In this direction, our work proposes two kinds of rules visualization. The first one is focused on visualization of one-to-one rules and the second one is focused on the n-to-n. In fact, for both visualizations, we have two filters options, one filter that can be applied before mining, it is useful for filtering the attributes that are supposed to be mined, and the other filter can be applied after the mining process to filtering rules. The visualization of spatiotemporal association rules calls for more complex interactive mechanisms in the attempt to explain the rules and its relationships once it can be any combination of n-to-n rules.
Findings- Association rule learning is a popular and well-researched set of methods for discovering interesting relations between entities in large databases in real-world problems. In this regard, an intelligent offshore oil industry environment is a very complex scenario and Occupational Health and Security (OHS) is a priority issue as it is an important factor to reduce the number of accidents and incidents records. In the oil industry, there exist standards to identify and record workplace accidents and incidents in order to provide guiding means on prevention efforts, indicating specific failures or reference, means of correction of conditions or circumstances that culminated in accident. In this paper, we propose a novel visualization of spatio-temporal rules that provides the big picture about risk analysis in a real world environment. Our main contribution lies in an interactive visualization for accident interpretation by means of well-defined spatio-temporal constraints in the offshore oil industry domain.
Research limitations/implications- Although we have introduced and presented the interactive visualization of association rules problem itself, it must be pointed out that, this approach is currently deployed as part of a larger system that rely of the mining and classification modules. In addition, a set of usability tests will complement this study. The global system is currently in use by a major petroleum industry conglomerate of Brazil.
Practical implications- Development of visualization of big data and especially for association rules lacks of having a method that evaluates its benefits and we should search for novel methods based on own techniques.
Keyword: data visualization, big data applications, decision support systems, oil and gas industry
Paper type: Research paper