Constraint Score: A new filter method for feature selection with pairwise constraints

Zhang, Daoqiang; Chen, Songcan; Zhou, Zhi‐Hua

doi:10.1016/j.patcog.2007.10.009

Cited by 184 publications

(102 citation statements)

References 14 publications

Supporting

Mentioning

101

Contrasting

Order By: Relevance

“…Filter-based methods [23] are independent from the classifier and focuses on the intrinsic properties of the original data. It usually provides a feature weighting or ranking based on some evaluation criteria and outputs a subset of selected features.…”

Section: Literature Reviewmentioning

confidence: 99%

P2P Lending Analysis Using the Most Relevant Graph-Based Features

Cui

Bai

Wang

et al. 2016

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. Peer-to-Peer (P2P) lending is an online platform to facilitate borrowing and investment transactions. A central problem for these P2P platforms is how to identify the most influential factors that are closely related to the credit risks. This problem is inherently complex due to the various forms of risks and the numerous influencing factors involved. Moreover, raw data of P2P lending are often high-dimension, highly correlated and unstable, making the problem more untractable by traditional statistical and machine learning approaches. To address these problems, we develop a novel filter-based feature selection method for P2P lending analysis. Unlike most traditional feature selection methods that use vectorial features, the proposed method is based on graph-based features and thus incorporates the relationships between pairwise feature samples into the feature selection process. Since the graph-based features are by nature completed weighted graphs, we use the steady state random walk to encapsulate the main characteristics of the graph-based features. Specifically, we compute a probability distribution of the walk visiting the vertices. Furthermore, we measure the discriminant power of each graph-based feature with respect to the target feature, through the Jensen-Shannon divergence measure between the probability distributions from the random walks. We select an optimal subset of features based on the most relevant graph-based features, through the Jensen-Shannon divergence measure. Unlike most existing state-of-the-art feature selection methods, the proposed method can accommodate both continuous and discrete target features. Experiments demonstrate the effectiveness and usefulness of the proposed feature selection algorithm on the problem of P2P lending platforms in China.

show abstract

Section: Literature Reviewmentioning

confidence: 99%

P2P Lending Analysis Using the Most Relevant Graph-Based Features

Cui

Bai

Wang

et al. 2016

Lecture Notes in Computer Science

View full text Add to dashboard Cite

show abstract

“…The highest fuzzy entropy value of the feature is regarded as the most informative one [14]. A feature f ∈ F is selected if it satisfies the following condition of Mean Selection (MS) Strategy as shown by equation (4).…”

Section: Feature Selectionmentioning

confidence: 99%

“…We have compared our proposed method results with other previous works in feature reduction of high dimensional data [1,4,7,8]. Table 4 shows the comparison of classification accuracy of our proposed method to other methods.…”

Section: % (6)mentioning

confidence: 99%

“…The filter approach was also usually performs some statistical analysis without employing any learning model. Zhang et al [4] have developed Constraint Score method. This is a filter method for feature selection with pairwise constraints, which specifies whether a pair of data samples belong to the same class (must-link constraints) or different classes (cannot-link constraint).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Untitled

Sari¹,

Muchtar²,

Zarkasi³

et al. 2014

Jurnal Ilmu Komputer dan Informasi

View full text Add to dashboard Cite

Curse of dimensionality is a major problem in most classification tasks. Feature transformation and feature selection as a feature reduction method can be applied to overcome this problem. Despite of its good performance, feature transformation is not easily interpretable because the physical meaning of the original features cannot be retrieved. On the other side, feature selection with its simple computational process is able to reduce unwanted features and visualize the data to facilitate data understanding. We propose a new feature selection method using similarity based entropy to overcome the high dimensional data problem. Using 6 datasets with high dimensional feature, we computed the similarity between feature vector and class vector. Then we find the maximum similarity that can be used for calculating the entropy values of each feature. The selected features are features that having higher entropy than mean entropy of overall features. The fuzzy k-NN classifier was implemented to evaluate the selected features. The experiment result shows that proposed method is able to deal with high dimensional data problem with mean accuracy of 80.5%. Keywords: classification, entropy, feature selection, high dimensional data, similarity AbstrakCurse of dimensionality merupakan masalah yang sering dihadapi pada proses klasifikasi. Transformasi fitur dan seleksi fitur sebagai metode dalam reduksi fitur bisa diterapkan untuk mengatasi masalah ini. Terlepas dari performanya yang baik, transformasi fitur sulit untuk diinterpretasikan karena ciri fisik dari fitur-fitur yang asli tidak dapat diperoleh kembali. Di sisi lain, seleksi fitur dengan proses komputasinya yang sederhana bisa mereduksi fitur-fitur yang tidak diperlukan dan mampu merepresentasikan data untuk memudahkan pemahaman terhadap data. Pada penelitian ini diajukan metode seleksi fitur baru yang berdasarkan pada dua pendekatan filter, yaitu similarity (kemiripan) dan entropi untuk mengatasi masalah data berdimensi tinggi. Tahap awal metode ini adalah menghitung nilai similarity antara fitur dengan vektor kelas dari 6 data berdimensi tinggi. Kemudian diperoleh nilai similarity maksimum yang digunakan untuk menghitung nilai entropi untuk setiap fitur. Fitur yang dipilih adalah fitur yang memiliki nilai entropi lebih tinggi daripada entropi rata-rata seluruh fitur. Fuzzy k-NN diterapkan untuk tahap klasifikasi data hasil seleksi fitur. Hasil percobaan menunjukkan bahwa metode yang diajukan mampu mengklasifikasi data berdimensi tinggi dengan rata-rata akurasi 80.5%.

show abstract

“…Spectral algorithms-based parewise constraints to learn distance metrics are stated by the authors in [20], [21]. Pairwise constraint is also used for feature selection [22]. Most existing semi-supervised clustering methods only ensure that the given constraints are considered, i.e., pairs of data points with ML (or CL) are not allocated to different clusters (or same clusters) or pairs of data points with ML (or CL) are close to (or distant from) each other in the transformed space or with the learned similarity measure.…”

mentioning

confidence: 99%

Constraint Neighborhood Projections for Semi-Supervised Clustering

Wang

et al. 2014

IEEE Trans. Cybern.

View full text Add to dashboard Cite

Abstract-Semi-supervised clustering aims to incorporate the known prior knowledge into the clustering algorithm. Pairwise constraints and constraint projections are two popular techniques in semi-supervised clustering. However, both of them only consider the given constraints and do not consider the neighbors around the data points constrained by the constraints. This paper presents a new technique by utilizing the constrained pairwise data points and their neighbors, denoted as constraint neighborhood projections that requires fewer labeled data points (constraints) and can naturally deal with constraint conflicts. It includes two steps: 1) the constraint neighbors are chosen according to the pairwise constraints and a given radius so that the pairwise constraint relationships can be extended to their neighbors, and 2) the original data points are projected into a new low-dimensional space learned from the pairwise constraints and their neighbors. A CNP-Kmeans algorithm is developed based on the constraint neighborhood projections. Extensive experiments on University of California Irvine (UCI) datasets demonstrate the effectiveness of the proposed method. Our study also shows that constraint neighborhood projections (CNP) has some favorable features compared with the previous techniques.Index Terms-Constraint neighborhood projections (CNP), pairwise constraints, semi-supervised clustering.

show abstract

Constraint Score: A new filter method for feature selection with pairwise constraints

Cited by 184 publications

References 14 publications

P2P Lending Analysis Using the Most Relevant Graph-Based Features

P2P Lending Analysis Using the Most Relevant Graph-Based Features

Untitled

Constraint Neighborhood Projections for Semi-Supervised Clustering

Contact Info

Product

Resources

About