F. Boylu

Bio: F. Boylu is an academic researcher from University of Florida. The author has contributed to research in topics: Stability (learning theory) & Online machine learning. The author has an hindex of 1, co-authored 1 publications receiving 9 citations.

Learning in the Presence of Self-Interested Agents

Abstract: In many situations a principal gathers a data sample containing positive and negative examples of a concept to induce a classification rule using a machine learning algorithm. Although learning algorithms differ from each other in various aspects, there is one essential issue that is common to all: the assumption that there is no strategic behavior inherent in the sample data generation process. In that respect, we ask the question "what if the observed attributes are being deliberately modified by the acts of some self-interested agents who will gain a preferred classification by engaging in such behavior". Therein such cases, there is a need for anticipating this kind of strategic behavior and incorporating it into the learning process. Classical learning approaches do not consider the existence of such behavior. In this paper we study the need for this kind of a paradigm and outline related research issues.

Inductive learning algorithms and representations for text categorization

Abstract: Text categorization-assignment of natural language texts to one or more predefined categories based on their content-is an important component in many information organization and management tasks.Different automatic learning algorithms for text categori-zation have different classification accuracy.Very accurate text classifiers can be learned automatically from training examples.

Induction over Strategic Agents

Abstract: We study the problem where a decision maker needs to discover a classification rule to classify intelligent, self-interested agents. Agents may engage in strategic behavior to alter their characteristics for a favorable classification. We show how the decision maker can induce a classification rule that anticipates such behavior while still satisfying an important risk minimization principle.

Principal-agent learning

Abstract: In this paper we present a merging, and hence an extension, of two recent learning methods, utility-based learning and strategic or adversarial learning. Recently, utility-based learning brings to the forefront the learner's utility function during induction. Strategic learning anticipates strategic activity in the induction process when the instances are intelligent agents such as in classification problems involving people or organizations. We call the resulting merged model principal-agent learning and present an induction process and example. Our model collapses to utility-based models when the agents do not engage in strategic behavior and to strategic learning when the learner's utility is not considered.

Enhanced ontology-based text classification algorithm for structurally organized documents

Abstract: Text classification (TC) is an important foundation of information retrieval and text mining. The main task of a TC is to predict the text‟s class according to the type of tag given in advance. Most TC algorithms used terms in representing the document which does not consider the relations among the terms. These algorithms represent documents in a space where every word is assumed to be a dimension. As a result such representations generate high dimensionality which gives a negative effect on the classification performance. The objectives of this thesis are to formulate algorithms for classifying text by creating suitable feature vector and reducing the dimension of data which will enhance the classification accuracy. This research combines the ontology and text representation for classification by developing five algorithms. The first and second algorithms namely Concept Feature Vector (CFV) and Structure Feature Vector (SFV), create feature vector to represent the document. The third algorithm is the Ontology Based Text Classification (OBTC) and is designed to reduce the dimensionality of training sets. The fourth and fifth algorithms, Concept Feature Vector_Text Classification (CFV_TC) and Structure Feature Vector_Text Classification (SFV_TC) classify the document to its related set of classes. These proposed algorithms were tested on five different scientific paper datasets downloaded from different digital libraries and repositories. Experimental obtained from the proposed algorithm, CFV_TC and SFV_TC shown better average results in terms of precision, recall, f-measure and accuracy compared against SVM and RSS approaches. The work in this study contributes to exploring the related document in information retrieval and text mining research by using ontology in TC.