A BIRCH-Based Clustering Method for Large Time Series Databases
Duong Pham Anh2012, Lecture Notes in Computer Science
visibility
…
description
2 pages
link
1 file
This paper presents a novel approach for time series clustering which is based on BIRCH algorithm. Our BIRCH-based approach performs clustering of time series data with a multi-resolution transform used as feature extraction technique. Our approach hinges on the use of cluster feature (CF) tree that helps to resolve the dilemma associated with the choices of initial centers and significantly improves the execution time and clustering quality. Our BIRCH-based approach not only takes full advantages of BIRCH algorithm in the capacity of handling large databases but also can be viewed as a flexible clustering framework in which we can apply any selected clustering algorithm in Phase 3 of the framework. Experimental results show that our proposed approach performs better than k-Means in terms of clustering quality and running time, and better than I-k-Means in terms of clustering quality with nearly the same running time.
Related papers
BIRCH: A New Data Clustering Algorithm and Its Applications
Data Mining and Knowledge Discovery, 1997
Data clustering is an important technique for exploratory data analysis, and has been studied for several years. It has been shown to be useful in many practical domains such as data classification and image processing. Recently, there has been a growing emphasis on exploratory analysis of very large datasets to discover useful patterns and/or correlations among attributes. This is called data mining, and data clustering is regarded as a particular branch. However existing data clustering methods do not adequately address the problem of processing large datasets with a limited amount of resources (e.g., memory and cpu cycles). So as the dataset size increases, they do not scale up well in terms of memory requirement, running time, and result quality.
Time Series Clustering Based on the K-Means Algorithm
Journal La Multiapp
Time series is one of the forms of data presentation that is used in many studies. It is convenient, easy and informative. Clustering is one of the tasks of data processing. Thus, the most relevant currently are methods for clustering time series. Clustering time series data aims to create clusters with high similarity within a cluster and low similarity between clusters. This work is devoted to clustering time series. Various methods of time series clustering are considered. Examples are given for real data.
BIRCH: an efficient data clustering method for very large databases
1996
Finding useful patterns in large datasets has attracted considerable interest recently, and one of the most widely st,udied problems in this area is the identification of clusters, or deusel y populated regions, in a multi-dir nensi onal clataset. Prior work does not adequately address the problem of large datasets and minimization of 1/0 costs.
A Novel Approach for High Dimensional Data Clustering
Clustering is considered as the most important unsupervised learning problem. It aims to find some structure in a collection of unlabeled data. Dealing with a large quantity of data items can be problematic because of time complexity. On the other hand high dimensional data is a challenge arena in data clustering e.g. time series data. Novel algorithms are needed to be robust, scalable, efficient and accurate to cluster of these kinds of data. In this study we proposed a two stages algorithm base on K-Means to achieve our objective.
A hybrid algorithm for clustering of time series data based on affinity search technique
TheScientificWorldJournal, 2014
Time series clustering is an important solution to various problems in numerous fields of research, including business, medical science, and finance. However, conventional clustering algorithms are not practical for time series data because they are essentially designed for static data. This impracticality results in poor clustering accuracy in several systems. In this paper, a new hybrid clustering algorithm is proposed based on the similarity in shape of time series data. Time series data are first grouped as subclusters based on similarity in time. The subclusters are then merged using the k-Medoids algorithm based on similarity in shape. This model has two contributions: (1) it is more accurate than other conventional and hybrid approaches and (2) it determines the similarity in shape among time series data with a low complexity. To evaluate the accuracy of the proposed model, the model is tested extensively using syntactic and real-world time series datasets.
Comprehensive analysis of different clustering algorithm and their impact over time series data
2015
Clustering is an approach to divide data into number of groups on the basis of some mutual characteristics each group called clusters, consists of objects that are similar between themselves and dissimilar to objects between other groups. Nowadays a lot of work is being carried out for trend analysis on time series data set. The genetic algorithm suffers with the problem in case of time series data because they consider each time stamp as a single entity during clustering. Even some algorithm still give good result on some type of time series data set but overall there is no generalized algorithm which considers different type of time series data set. This paper is presents comprehensive analysis over different type of clustering algorithms like k-means, hierarchical, SOM and GMM on two type of time series data, microarray and financial. These algorithms are compared on the following factors: size of data, number of clusters, type of data set, homogeneity score, separation score, si...
Characteristic-Based Clustering for Time Series Data
Data Mining and Knowledge Discovery, 2006
Improved Multi Threshold Birch Clustering Algorithm
BIRCH algorithm is a clustering algorithm suitable for very large data sets. In the algorithm, a CF-tree is built whose all entries in each leaf node must satisfy a uniform threshold T, and the CF-tree is rebuilt at each stage by different threshold. But using a single threshold cause many shortcomings in the birch algorithm, in this paper to propose a solution to this shortcoming by using multiple thresholds instead of a single threshold.
Dimension reduction for clustering time series using global characteristics
2005
Existing methods for time series clustering rely on the actual data values can become impractical since the methods do not easily handle dataset with high dimensionality, missing value, or different lengths. In this paper, a dimension reduction method is proposed that replaces the raw data with some global measures of time series characteristics. These measures are then clustered using a self-organizing map.
SEGMENTING TIME SERIES: A SURVEY AND NOVEL APPROACH
Series in Machine Perception and Artificial Intelligence, 2004
In recent years, there has been an explosion of interest in mining time series databases. As with most computer science problems, representation of the data is the key to efficient and effective solutions. One of the most commonly used representations is piecewise linear approximation. This representation has been used by various researchers to support clustering, classification, indexing and association rule mining of time series data. A variety of algorithms have been proposed to obtain this representation, with several algorithms having been independently rediscovered several times. In this paper, we undertake the first extensive review and empirical comparison of all proposed techniques. We show that all these algorithms have fatal flaws from a data mining perspective. We introduce a novel algorithm that we empirically show to be superior to all others in the literature.
A BIRCH-Based Clustering Method
for Large Time Series Databases
Vo Le Quy Nhon and Duong Tuan Anh
Faculty of Computer Science and Engineering,
Ho Chi Minh City University of Technology, Vietnam
[email protected]
Abstract. This paper presents a novel approach for time series clustering which is based on BIRCH algorithm. Our BIRCH-based approach
performs clustering of time series data with a multi-resolution transform
used as feature extraction technique. Our approach hinges on the use of
cluster feature (CF) tree that helps to resolve the dilemma associated
with the choices of initial centers and significantly improves the execution time and clustering quality. Our BIRCH-based approach not only
takes full advantages of BIRCH algorithm in the capacity of handling
large databases but also can be viewed as a flexible clustering framework
in which we can apply any selected clustering algorithm in Phase 3 of
the framework. Experimental results show that our proposed approach
performs better than k-Means in terms of clustering quality and running time, and better than I-k-Means in terms of clustering quality with
nearly the same running time.
Keywords: time series clustering, cluster feature, DWT, BIRCH-based.
1
Introduction
Time series data arise in so many applications of various areas ranging from science, engineering, business, finance, economic, medicine to government. Because
of this fact, there has been an explosion of research effort devoted to time series
data mining in the last decade, in particular, in behavior-related data analytics
[8]. Beside similarity search, another crucial task in time series data mining which
has received an increasing amount of attention lately is time series clustering.
Given a set of unlabeled time series, it is often desirable to determine groups of
similar time series in such a way that time series belonging to the same group
are more “similar” to each other rather than time series from different groups.
Although there have been much research on clustering in general, most classic
machine learning and data mining algorithms do not work well for time series
due to their unique characteristics. In particular, the high dimensionality, very
high feature correlation and the large amount of noise in time series data present
a difficult challenge. Although a few time series clustering algorithms have been
proposed, most of them do not scale well to large datasets and work only in a
batch fashion.
L. Cao et al. (Eds.): PAKDD 2011 Workshops, LNAI 7104, pp. 148–159, 2012.
c Springer-Verlag Berlin Heidelberg 2012
A BIRCH-Based Clustering Method for Large Time Series Databases
149
This paper proposes a novel approach for time series clustering which is based
on BIRCH algorithm [13]. We adopt BIRCH method in the context of time series
data due to its three inherent benefits. First, BIRCH incrementally and dynamically clusters incoming multi-dimensional metric data points to try to produce
the best quality clustering with the available resources (i.e. available memory
and time constraints). Second, BIRCH can find a good clustering with a single
scan of the data and improve the quality further with a few additional scans.
Third, BIRCH can scale well to very large datasets. To deal with the characteristics of time series data, we propose a BIRCH-based clustering approach that
works by first using a multi-resolution transform to perform feature extraction
on all time series in the database and then applying BIRCH algorithm to cluster
the transformed time series data. Our BIRCH-based approach hinges on the use
of cluster feature (CF) tree that helps to resolve the dilemma associated with
the choices of initial centers and thus significantly improves the execution time
and clustering quality. Our BIRCH-based approach not only takes full advantages of BIRCH method in the capacity of handling very large databases but
also can be viewed as a flexible clustering framework in which we can apply
any selected clustering algorithm in Phase 3 of the framework. Particularly, our
BIRCH-based approach with an anytime clustering algorithm used in Phase 3
will also be an anytime algorithm. Experimental results show that our proposed
approach performs better than k-Means in terms of clustering quality and running time, and better than I-k-Means in terms of clustering quality with nearly
the same running time.
The rest of the paper is organized as follows. In Section 2 we review related
work, and introduce the necessary background on multi-resolution transforms
and BIRCH clustering algorithm. Section 3 describes our proposed approach
for time series clustering. Section 4 presents our experimental evaluation which
compares our method to classic k-Means and I-k-Means on real datasets. In
section 5 we include some conclusions and suggestions for future work.
2
2.1
Background and Related Work
Dimensionality Reduction Using Multi-resolution Transforms
Discrete wavelet transform (DWT) is a typical case of time series dimensionality
reduction method with multi-resolution property. This property is critical to
our proposed framework. The Haar wavelet is the simplest and most popular
wavelet proposed by Haar. The Haar wavelet transform can be seen as a sequence
of averaging and difference operations on a time sequence. The transform is
achieved by averaging two adjacent values on the time series at a given resolution
to form a smoothed, lower-dimensional signal and the resulting coefficients at
this given resolution are simply the differences between the values and their
averages [1].
MPAA time series representation, proposed by Lin et al., 2005 [7], is another
case of dimensionality reduction with multi-resolution property. MPAA divides
time series Xi of length n into a series of lower dimensional signal with different
Related papers
O PROFESSOR DIANTE DA DEMANDA DO ALUNO DO SÉCULO XXI: TRABALHANDO COM TECNOLOGIAS E MÍDIAS DE POTENCIAL EDUCATIVO
Debates em Educação, 2019
İklim Değişikliğine Bağlı Göçlerin Avrupa Birliği'ne Üye Ülkeler ve Türkiye'de Sosyal Güvenlik Sistemlerine Etkileri
İklim Değişikliğine Bağlı Göçlerin Avrupa Birliği'ne Üye Ülkeler ve Türkiye'de Sosyal Güvenlik Sistemlerine Etkileri, 2019
Effect of Organic Carbon and Nitrogen on the Interactions of Morchella spp. and Bacteria Dispersing on Their Mycelium
Frontiers in Microbiology, 2019
A Wavelet-Based OFDM System Implemetation on GNURadio Platform vs. an FFT-Based
Communications in Computer and Information Science, 2017
Basal implants- A new era of prosthodontic dentistry
IP Annals of Prosthodontics and Restorative Dentistry, 2020
Examining leadership skills, behaviors, and effective communication for virtual IT project managers
Team Performance Management, 2022
Nuevas prácticas escénicas y dramáticas: fenómenos parateatrales, no teatrales o antiteatrales
Desde el Sur. Revista de Ciencias Humanas y Sociales de la Universidad Científica del Sur, 2013
Multiple Measurements of Quasars Acting as Standard Probes: Exploring the Cosmic Distance Duality Relation at Higher Redshift
The Astrophysical Journal, 2020
El cumplimiento de prestaciones dinerarias en el Código Civil peruano
Thēmis Revista De Derecho, 2015