The problem of finding interesting and actionable patterns is a major challenge in data mining. It has been studied by many data mining researchers. The issue is that data mining algorithms often generate too many patterns, which make it very hard for the user to find those truly useful ones. Over the years many techniques have been proposed. However, few have made it to real-life applications. At the end of 2005, we built a data mining system for Motorola (called opportunity map) to enable the user to explore the space of a large number of rules in order to find actionable knowledge. The approach is based on the concept of rule cubes and operations on rule cubes. A rule cube is similar to a data cube, but stores rules. Since its deployment, some issues have also been identified during the regular use of the system in Motorola. One of the key issues is that although the operations on rule cubes are flexible, each operation is primitive and has to be initiated by the user. Finding a piece of actionable knowledge typically involves many operations and intense visual inspections, which are labor-intensive and time-consuming. From interactions with our users, we identified a generic problem that is crucial for finding actionable knowledge. The problem involves extensive comparison of sub-populations and identification of the cause of their differences. This paper first defines the problem and then proposes an effective method to solve the problem automatically. To the best of our knowledge, there is no reported study of this problem. The new method has been added to the opportunity map system and is now in daily use in Motorola.

The problem of interestingness of discovered rules has been investigated by many researchers. The issue is that data mining algorithms often generate too many rules, which make it very hard for the user to find the interesting ones. Over the years many techniques have been proposed. However, few have made it to real-life applications. Since August 2004, we have been working on a major application for Motorola. The objective is to find causes of cellular phone call failures from a large amount of usage log data. Class association rules have been shown to be suitable for this type of diagnostic data mining application. We were also able to put several existing interestingness methods to the test, which revealed some major shortcomings. One of the main problems is that most existing methods treat rules individually. However, we discovered that users seldom regard a single rule to be interesting by itself. A rule is only interesting in the context of some other rules. Furthermore, in many cases, each individual rule may not be interesting, but a group of them together can represent an important piece of knowledge. This led us to discover a deficiency of the current rule mining paradigm. Using non-zero minimum support and non-zero minimum confidence eliminates a large amount of context information, which makes rule analysis difficult. This paper proposes a novel approach to deal with all of these issues, which casts rule analysis as OLAP operations and general impression mining. This approach enables the user to explore the knowledge space to find useful knowledge easily and systematically. It also provides a natural framework for visualization. As an evidence of its effectiveness, our system, called Opportunity Map, based on these ideas has been deployed, and it is in daily use in Motorola for finding actionable knowledge from its engineering and other types of data sets.

In this paper, we report a deployed data mining application system for Motorola. Originally, its intended use was for identifying causes of cellular phone failures, but it has been found to be useful for many other engineering data sets as well. For this report, the case study is a dataset containing cellular phone call records. This data set is like any dataset used in classification applications, i.e., with a set of attributes which can be continuous or discrete, and a discrete class attribute. In our application, the classes are normally ended calls, calls which failed to setup, and calls which failed while in progress. However, the task is not to predict any failure, but to identify possible causes that resulted in failures. Then, engineering efforts may focus on improvements that can be made to the phones. In the course of the project, various classification techniques, e.g., decision trees, naïve Bayesian classification and SVM were tried. However, the results were unsatisfactory. After several demonstrations and interaction with domain experts, we finally designed and implemented an effective approach to perform the task. The final system is based on class association rules, general impressions and visualization. The system has been deployed and is in regular use at Motorola. In this paper, we first describe our experiences with some existing classification systems and discuss why they are not suitable for the task. We then present our techniques. As an illustration, we show several visualization screens in the case study, which reveal some important knowledge. Due to confidentiality, we will not give specifics but only present a general discussion about the results.