Alev Mutlu

Improving the scalability of ILP-based multi-relational concept discovery system through parallelization

Due to the increase in the amount of relational data that is being collected and the limitations of propositional problem definition in relational domains, multi-relational data mining has arisen to be able to extract patterns from relational data. In order to cope with intractably large search space and still to be able to generate high-quality patterns, ILP-based multi-relational data mining and concept discovery systems employ several search strategies and pattern limitations. Another direction to cope with the large search space is using parallelization. By parallel data mining, improvement in time efficiency and scalability can be provided without further limiting the language patterns. In this work, we describe a method for concept discovery with parallelization on an ILP-based concept discovery system. The non-parallel algorithm, namely Concept Rule Induction System (CRIS), is modified in such a way that the parts that involve high amount of query processing, which causes bottleneck, are reorganized in a data parallel way. The resulting algorithm is called, Parallel CRIS (pCRIS). A set of experiments is conducted in order to evaluate the performance of the proposed method.

A Hybrid Graph-Based Method for Concept Rule Discovery

In this paper we present a hybrid graph-based concept discovery method. Concept rule discovery aims at finding the definition of a specific concept in terms of relations involving background knowledge. The proposed method is a hybrid approach that combines path finding-based and substructure-based approaches. It distinguishes from other state of the art graph-based approaches in such a way that the data is initially stored in a relational database, and the concept descriptors are induced while the graph is constructed. In this approach, in addition to being the representation framework, the graph structure is also utilized to guide the concept induction process. A set of experiments is conducted on data sets that belong to different learning problems. The results show that the proposed approach has promising results in comparison to state of the art methods in terms of running time and accuracy.

A Graph-Based Concept Discovery Method for n-Ary Relations

Concept discovery is a multi-relational data mining task for inducing definitions of a specific relation in terms of other relations in the data set. Such learning tasks usually have to deal with large search spaces and hence have efficiency and scalability issues. In this paper, we present a hybrid approach that combines association rule mining methods and graph-based approaches to cope with these issues. The proposed method inputs the data in relational format, converts it into a graph representation, and traverses the graph to find the concept descriptors. Graph traversal and pruning are guided based on association rule mining techniques. The proposed method distinguishes from the state-of-the art methods as it can work on n-ary relations, it uses path finding queries to extract concepts and can handle numeric values. Experimental results show that the method is superior to the state-of-the art methods in terms of accuracy and the coverage of the induced concept descriptors and the running time.

Improving the Time Efficiency of ILP-based Multi-Relational Concept Discovery with Dynamic Programming Approach

Large amount of relational data is stored in databases. There- fore, working directly on the data stored in database is an important feature for multi-relational concept discovery systems. In addition to concept rule quality, time efficiency is an important performance dimension for concept discovery since dealing with large amount of data is a must. In this work, we present a dynamic programming based approach for improving the time efficiency on an ILP-based concept discovery system, namely CRIS (Concept Rule Induction System), which combines ILP and Apriori and directly works on databases.

A path-finding based method for concept discovery in graphs

In the multi-relational data mining, concept discovery is the problem of inducing definitions of a relation in terms of other relations provided. In this paper, we present a method that combines graph-based and association rule mining-based methods for concept discovery in graphs. The proposed method is related to graphs as the data, which is initially stored in a relational database, is represented as a graph and concept descriptors are the paths that connect certain vertices; and it is related to association rule mining as it uses methods of association rule mining to prune the search space and evaluate the quality of the concept descriptors. The method is evaluated on several data sets, and the experimental results show that it is compatible with the state-of-the art methods in terms of accuracy and coverage of the induced concept descriptors and the running time of the application.

Improving Hash Table Hit Ratio of an ILP-Based Concept Discovery System with Memoization Capabilities

Although ILP-based concept discovery systems have applications in a wide range of domains, they still suffer from scalability and efficiency issues. One of the reasons for the efficiency problem is high number of query executions necessary in the concept discovery process. Due to refinement operator of ILP-based systems, these queries repeat frequently. In this work we propose a method to improve hash table hit ratio for repeating queries of ILP-based concept discovery systems with memoization capabilities. The proposed method introduces modifications on search space evaluation and covering steps of such systems. Experimental results show that the proposed method improves the hash table hit count of ILP-based concept discovery systems with an affordable cost of extra memory consumption.

Personalized time-aware outdoor activity recommendation system

Location-based social network as one of the platforms in this field has been providing services and facilities to enhance user experience to explore their surroundings and new places. Among current services, point of interest (POI) recommendation and activity recommendation draws significant attention of users, which makes it a potential field of the study. However, despite of all the developments performed in this field, activity recommendation system still requires further improvements, since only a few related studies concentrated on this topic so far. In order to develop the activity recommendation system, in this work two approaches are presented, using different ideas by extending existing location-based collaborative filtering (CF) recommendation models. One of them focuses on the temporal feature of the data and the other one emphasizes on the correlation between activities to estimate the probability of selecting each activity. The proposed methods are evaluated on a medium-scale real data set obtained by combining the Gowalla and Foursqaure. The experimental results confirm that both methods remarkably outperform the basic CF model.

Concept Discovery in Graph Databases

Concept discovery is one of the most commonly addressed tasks of multi-relational data mining and is concerned with inducing logical definitions of a relation in terms of other relations provided. The problem has long been studied from Inductive Logic Programming and graph-oriented perspectives. In this study, we investigate the problem from graph databases perspective and propose a pathfinding-based method for concept discovery in graph databases. More specifically, we introduce a method that employs Neo4j graph database technology to store data and find the concept descriptors that define the target relation and implements several techniques to further improve the post processing steps of concept discovery process. The experimental results show that the proposed method is superior to state-of-the art concept discovery systems in terms of rule induction time, discovers shorter concept descriptors with high coverage and F1 score, and scales well.

A Graph-Path Counting Approach for Learning Head Output Connected Relations

Concept discovery is a multi-relational data mining task where the problem is inducing definitions of a relation in terms of other relations. In this paper, we propose a graph-based concept discovery system to learn definitions of head output connected relations. It inputs the data in relational format, converts it into graphs, and induces concept definitions from graphs’ paths. The proposed method can handle n-ary relations and induce recursive concept definitions. Path frequencies are used to calculate the quality of the induced concept descriptors. The experimental results show that results obtained are comparable to those reported in literature in terms of running time and coverage; and is superior over some methods as it can induce shorter concept descriptors with the same coverage.

Graph-based concept discovery in multi relational data

Developments in technology, especially in computer science created the need of storing data in variety of areas. This need created the term database where the data is stored in a useful form. In the database, data is logically integrated in file/files according to relations among them. One of the important issues is to extract knowledge from these databases that hold data in a useful and complete form. This process is called as data mining. The main objective of data mining is to extract implicit and useful knowledge from huge and at first glance meaningless mass of data that is stored in database(s). Multi-Relational databases are the ones in which the data is stored in multiple tables (relations). The relationships between those tables are also stored as tables (relations) in the database. The more effective and commonly known approaches for Multi-Relational Data Mining (MRDM) are based on Inductive Logic Programming (ILP). ILP contains concepts from Inductive Learning and Logic Programming. From this point, the main purpose of MRDM is extracting implicit and trivial knowledge from relational database(s) using ILP approaches and techniques. In this approach, data is represented in graph structures and graph mining techniques are used for knowledge discovery. Concept discovery in multi-relational data mining aims to find relational rules that best describe a relation, called target relation, in terms of other relations in the database, called background knowledge. In this study, a graph-based concept discovery method for concept discovery is presented. The proposed method, namely G-CDS (Graph-based Concept Discovery System), utilizes methods both from substructure-based and path-finding based approaches, hence it can be considered as a hybrid method. G-CDS generates disconnected graph structures for each target relation and its related background knowledge, which are initially stored in a relational database, and utilizes them to guide generation of a summary graph. The summary graph is traversed to find concept descriptors. A set of experiments is conducted on datasets that belong to different learning problems. The experimental results show that G-CDS is capable of learning definitions of target relations that belong to different learning problems.