Personalized Meta Actions

Personalized Meta Actions

Hakim Touati (University of North Carolina at Charlotte, USA) and Zbigniew W. Ras (University of North Carolina at Charlotte, USA, & Warsaw University of Technology, Poland)
Copyright: © 2014 |Pages: 10
DOI: 10.4018/978-1-4666-5202-6.ch166

Chapter Preview



Recently, there has been an increasing interest in business analytics and big data tools to understand and drive industries evolution. Today’s corporations are interested in understanding their customers and keeping them loyal. In fact keeping a loyal customer cost seven times less then getting a new one1. They are interested in understanding their manufacturing process and decreasing their costs. Distribution companies strive to optimize their distribution costs and improve their revenue. The healthcare industry is also interested in new methods to analyze data and provide better care. Given the wealth of data that corporations are accumulating, it is natural to take advantage of data driven decision-making solutions. Researchers proposed solutions ranging from data mining and pattern recognition to forecasting new trends. One of the most promising techniques in this field is action rule discovery.

Action rule discovery is a new efficient technique that drives evolution in desirable directions. Action rules are already adopted by the banking industry, distribution industry, and healthcare industry to discover the right actions and strategies to follow in order to increase their profit. For instance, (Wasyluk, Ras, & Wyrzykowska, 2008; Zhang, Ras, Jastreboff, & Thompson, 2010) studied action rules in the healthcare domain to improve patient’s care.

Action rules can be further improved by the introduction of meta-actions that help corporations control their actions. Meta-actions trigger action rules in order to let corporations control their investment cost and return on investment before hand; thus, constructing their strategies. Depending on the meta-actions used.

Action rules are mined on entire objects’ or customers’ populations, or all aspects related to certain processes. Meta-actions, on the other hand, are chosen based on the action rules. However, applying those techniques blindly without studying the effects on customers or objects might result in negative side effects. Depending on the industry and the objects having actions upon them, an object centric approach is the most suitable approach to answer personalized needs of specific objects. Therefore, we felt the need to introduce personalization on meta-actions when executing action rules. Personalization is a very important aspect in driving evolution, and should be part of corporations’ strategy. In this chapter, we propose the first personalized meta-action selection technique that takes into consideration the possible negative side effects on customers when applying such meta-actions. We further strive to optimize the cost and weight (features importance) effects of applying meta-actions on any action rule strategy for specific customers. In addition, we propose a more relaxed strategy to group patients and meta-actions based on negative side effects.

This chapter’s contributions are as follow:

  • 1.

    Optimizing the meta-actions selection by optimizing cost / weight effect.

  • 2.

    Proposing a grouping mechanism for objects and meta-actions.

  • 3.

    Providing an incremental study analysis.

We start by introducing some preliminaries, and visit the related work. We then define the problem and challenges related to personalization. We describe our proposed approach in section 4, and the incremental behavior of the approach in section 5. We conclude and discuss the contributions and shortcomings of our approach in section 6.



Features are attributes describing objects’ properties (i.e. customers properties), and are recorded in a database as transactions. For instance, a bank’s customer might be described by his mortgage rate, his salary, and his address as features stored in one database transaction. A hospital’s patient temperature, blood pressure, and age might be observed during his medical examination and stored in his electronic medical record as features. In addition, features are labeled with weights that represent their importance among other features. Experts knowledge need to be introduced in the system in order to label the features with weights, otherwise the weight is normalized to one.

Key Terms in this Chapter

Rule Support: The number of objects supporting a rule. In action rules, it is the number of objects supporting the antecedent side of the rule

Side Effect: Secondary, usually undesirable effect resulting from applying a meta-action and affecting the objects negatively. For instance, medical treatments result in side effects.

Cold Start: Potential problem involving insufficient data to draw an inference on specific objects. It usually occurs when systems are first deployed with a lack of data

Action Rule: Rules providing action leading in transitioning an object to a more profitable state using multiple flexible features transformation.

Flexible Features: Object properties that can transition from one value to another value triggering a change in the object state.

Meta-Action: An action resulting in a set of atomic action terms triggering an action rule. For instance, taking a drug is a meta-action.

Complete Chapter List

Search this Book: