Positive Unlabeled Learning for Data Stream Classification

Li, Xiao Li; Yu, Philip S.; Liu, Bing; Ng, See Ket

doi:10.1137/1.9781611972795.23

Cited by 100 publications

(85 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The proposed one-pass incremental change detection algorithm was applied to the detection on video-shot changes. Li et al (2009b) proposed a novel PU (Positive and Unlabelled examples) learning technique LELC (PU Learning by Extracting Likely positive and negative microClusters). It only requires a small set of positive examples and a set of unlabelled examples to build accurate classifiers.…”

Section: Literature Reviewmentioning

confidence: 99%

Adaptive building of hybrid learning model for detecting and adapting concept drifting data streams

Patil

Kulkarni

2014

IJKEDM

View full text Add to dashboard Cite

Abstract:In recent developments in electricity market deregulation, the prices are not fixed. In such application, class labels are not available directly and potentially valuable information is lost. A learning model of electricity demand and prices needs to be adaptive for dynamic changes in massive data streams. This paper presents adaptive building of hybrid learning model for electricity prices by detecting and adapting concept drifting data streams. A proposed framework is build with four main challenges such as future assumption, change type, learner adaptivity and model. A decision tree is built incrementally using Q-learning and Hoeffding bounds. A k-mode method produces concept clusters to handle unlabelled data, change type in terms of deviations between past concept clusters and current ones and predicts future assumptions. An adaptive classify algorithm is developed for the predictive ability evaluation on the test set. Result of experiments using Elec2 data confirms applicability of methodology with more than 80% unlabelled data.Keywords: adaptive learning model; decision tree; data streams; concept drift; reinforcement learning.Reference to this paper should be made as follows: Kumar, S., Singh, V. and Haleem, A. (2014)

show abstract

Section: Literature Reviewmentioning

confidence: 99%

Adaptive building of hybrid learning model for detecting and adapting concept drifting data streams

Patil

Kulkarni

2014

IJKEDM

View full text Add to dashboard Cite

show abstract

“…For the methods in the first category [26,27,28], they are developed mainly for document-related one-class classification problems. Their main task is to extract negative samples from unlabeled examples (if unlabeled examples are offered) and to construct a binary classifier using target samples and extracted negative samples.…”

Section: One-class Learningmentioning

confidence: 99%

One-Class-Based Uncertain Data Stream Learning

Liu

Xiao

Cao

et al. 2011

Proceedings of the 2011 SIAM International Conference on Data Mining

Self Cite

View full text Add to dashboard Cite

This paper presents a novel approach to one-class-based uncertain data stream learning. Our proposed approach works in three steps. Firstly, we put forward a local kerneldensity-based method to generate a bound score for each instance, which refines the location of the corresponding instance. Secondly, we construct an uncertain one-class classifier by incorporating the generated bound score into a one-class SVM-based learning phase. Thirdly, we devise an ensemble classifier, integrated from uncertain one-class classifiers built on the current and historical chunks, to cope with the concept drift involved in the uncertain data stream environment. Our proposed method explicitly handles the uncertainty of the input data and enhances the ability of oneclass learning in reducing the sensitivity to noise. Extensive experiments on uncertain data streams demonstrate that our proposed approach can achieve better performance and is highly robust to noise in comparison with state-of-the-art one-class learning method.

show abstract

“…This setting is also known as Positive and Unlabeled (PU) learning, which aims to learn from data with only positive and unlabeled examples. It has been shown to be useful in many data mining tasks [8], [9], [10], [11], such as text mining, uncertain data mining, stream mining, etc. Formally, the graph PU learning problem corresponds to training a model to identify a subset of unlabeled graphs that are most likely to be negative graphs.…”

Section: Introductionmentioning

confidence: 99%

“…Conventional PU learning approaches can identify a group of reliable negative examples in the vector space. These approaches assume that the full set of useful features is available [9], [10], [11]. However, graph data are not directly represented in a feature space, and they require an additional subgraph feature mining process by evaluating the subgraph patterns in a graph data set.…”

Section: Introductionmentioning

confidence: 99%

Positive and Unlabeled Learning for Graph Classification

Zhao

Kong

2011

2011 IEEE 11th International Conference on Data Mining

Self Cite

View full text Add to dashboard Cite

Abstract-The problem of graph classification has drawn much attention in the last decade. Conventional approaches on graph classification focus on mining discriminative subgraph features under supervised settings. The feature selection strategies strictly follow the assumption that both positive and negative graphs exist. However, in many real-world applications, the negative graph examples are not available. In this paper we study the problem of how to select useful subgraph features and perform graph classification based upon only positive and unlabeled graphs. This problem is challenging and different from previous works on PU learning, because there are no predefined features in graph data. Moreover, the subgraph enumeration problem is NP-hard. We need to identify a subset of unlabeled graphs that are most likely to be negative graphs. However, the negative graph selection problem and the subgraph feature selection problem are correlated. Before the reliable negative graphs can be resolved, we need to have a set of useful subgraph features. In order to address this problem, we first derive an evaluation criterion to estimate the dependency between subgraph features and class labels based on a set of estimated negative graphs. In order to build accurate models for the PU learning problem on graph data, we propose an integrated approach to concurrently select the discriminative features and the negative graphs in an iterative manner. Experimental results illustrate the effectiveness and efficiency of the proposed method.

show abstract

Positive Unlabeled Learning for Data Stream Classification

Cited by 100 publications

References 30 publications

Adaptive building of hybrid learning model for detecting and adapting concept drifting data streams

Adaptive building of hybrid learning model for detecting and adapting concept drifting data streams

One-Class-Based Uncertain Data Stream Learning

Positive and Unlabeled Learning for Graph Classification

Contact Info

Product

Resources

About