Terms Of F1 Research Articles

Multitask defect prediction

AbstractWithin‐project defect prediction assumes that we have sufficient labeled data from the same project, while cross‐project defect prediction assumes that we have plenty of labeled data from source projects. However, in practice, we might only have limited labeled data from both the source and target projects in some scenarios. In this paper, we want to apply multitask learning to investigate such a new scenario. To our best knowledge, this problem (ie, both the source project and the target project have limited labeled data) has not been thoroughly investigated, and we are the first to propose a novel multitask defect prediction approach mask. mask consists of a differential evolution optimization phase and a multitask learning phase. The former phase aims to find optimal weights for shared and nonshared information in related projects (ie, the target project and its related source projects), while the latter phase builds prediction models for each project simultaneously. To verify the effectiveness of mask, we perform experimental studies on 18 real‐world software projects and compare our approach with four state‐of‐the‐art baseline approaches: single‐task learning (STL), simple combined learning (SCL), Peters filter, and Burak filter. Experimental results show that mask can achieve F1 of 0.397 and AUC of 0.608 on average with a few labeled data (ie, 10% of data). Across the 18 projects, mask can outperform baseline methods significantly in terms of F1 and AUC. Therefore, by utilizing the relatedness among multiple projects, mask can perform significantly better than the state‐of‐the‐art methods. The results confirm that mask is promising for software defect prediction when the source and target projects both have limited training data.

Mining Fraudsters and Fraudulent Strategies in Large-Scale Mobile Social Networks

The rapid development of modern communication technologies-in particular, (mobile) phone communications-has largely facilitated human social interactions and information exchange. However, the emergence of telemarketing frauds can significantly dissipate individual fortune and social wealth, resulting in a potential slow down or damage to economics. In this work, we propose to spot telemarketing frauds, with an emphasis on unveiling the “precise fraud” phenomenon and the strategies that are used by fraudsters to precisely select targets. To study this problem, we employ a one-month complete dataset of telecommunication metadata in Shanghai with 54 million users and 698 million call logs. Through our study, we find that user's information might have been seriously leaked, and fraudsters have a preference over the target user's age and activity in mobile network. We further propose a novel semi-supervised learning framework to distinguish fraudsters from non-fraudsters. Experimental results on a real-world data show that our approach outperforms several state-of-the-art algorithms in accuracy of detecting fraudsters (e.g., +0.278 in terms of F1 on average). We believe that our study can potentially inform policymaking for government and mobile service providers.

Factor Graph Model Based User Profile Matching Across Social Networks

In modern society, it is common for people to be active in many different online social networks at once. As new social network services arise every year, it remains a great challenge to integrate social data. Discovering multiple profiles of a single person across different social networks is a precondition for integration, but it is still challenging due to the inconsistency and disruption of the accessible information among social media networks (SMNs). Many studies have made efforts on user's profiles, users' contents, and network structure to address this issue, but the issue of how to consider all these information in a unified model and tackle them simultaneously still remains challenging. Considering that identical users tend to have partial similar friend relationship structures in different SMNs, especially friendship SMNs, we deepen the analysis of “friend” relationships (mutual following connections) in different SMNs, and propose PIFGM (Pairwise Identical Factor Graph Model), a novel factor graph model-based model, to address this problem by considering both user attributes and friend relationships across networks. We also present a distributed learning algorithm to handle large-scale social networks. We evaluate the proposed model on two different data collections: SNS and SR. Our experimental results validate the effectiveness and efficiency of the proposed model. The proposed PIFGM significantly outperforms several alternative methods by up to approximately 10%~20% in terms of F1 and precision on SNS and SR respectively.

A new sequence based encoding for prediction of host–pathogen protein interactions

Pathogen–host interactions are very important to figure out the infection process at the molecular level, where pathogen proteins physically bind to human proteins to manipulate critical biological processes in the host cell. Data scarcity and data unavailability are two major problems for computational approaches in the prediction of pathogen–host interactions. Developing a computational method to predict pathogen–host interactions with high accuracy, based on protein sequences alone, is of great importance because it can eliminate these problems. In this study, we propose a novel and robust sequence based feature extraction method, named Location Based Encoding, to predict pathogen–host interactions with machine learning based algorithms. In this context, we use Bacillus Anthracis and Yersinia Pestis data sets as the pathogen organisms and human proteins as the host model to compare our method with sequence based protein encoding methods, which are widely used in the literature, namely amino acid composition, amino acid pair, and conjoint triad. We use these encoding methods with decision trees (Random Forest, j48), statistical (Bayesian Networks, Naive Bayes), and instance based (kNN) classifiers to predict pathogen–host interactions. We conduct different experiments to evaluate the effectiveness of our method. We obtain the best results among all the experiments with RF classifier in terms of F1, accuracy, MCC, and AUC.

Predicting Effectiveness of Generate-and-Validate Patch Generation Systems Using Random Forest

One way to improve practicability of automatic program repair (APR) techniques is to build prediction models which can predict whether an application of a APR technique on a bug is effective or not. Existing prediction models have some limitations. First, the prediction models are built with hand crafted features which usually fail to capture the semantic characteristics of program repair task. Second, the performance of the prediction models is only evaluated on Genprog, a genetic-programming based APR technique. This paper develops prediction models, i.e., random forest prediction models for SPR, another kind of generate- and-validate APR technique, which can distinguish ineffective repair instances from effective repair instances. Rather than handcrafted features, we use features automatically learned by deep belief network (DBN) to train the prediction models. The empirical results show that compared to the baseline models, that is, all effective models, our proposed models can at least improve the F1 by 9% and AUC(area under the receiver operating characteristics curve) by 19%. At the same time, the prediction model using learned features at least outperforms the one using hand-crafted features in terms of F1 by 11%

A novel sentence similarity model with word embedding based on convolutional neural network

SummaryIn this paper, we propose an effective model for the similarity metrics of English sentences. In the model, we first make use of word embedding and convolutional neural network (CNN) to produce a sentence vector and then leverage the information of the sentence vector pair to calculate the score of sentence similarity. Considering the case of long‐range semantic dependencies between words, we propose a novel method transforming word embeddings to construct the three‐dimensional sentence feature tensor. In addition, we incorporate the k‐max pooling into the convolutional neural network to adapt to variable lengths of input sentences. The proposed model requires no external resource such as WordNet and parse tree. Meanwhile, it consumes very little time for training. Finally, we carried out extensive simulations to evaluate the performance of our model compared with other state‐of‐the‐art works. Experimental results on SemEval 2014 task (SICK test corpus) indicated that our model can achieve a good performance in the terms of Pearson correlation coefficient, Spearman correlation coefficient, and mean squared errors. Furthermore, experimental results on Microsoft research paraphrase identification (MSRP) indicated that our model can achieve an excellent performance in the terms of F1 and Accuracy.

A benchmark study on the effectiveness of search-based data selection and feature selection for cross project defect prediction

ContextPrevious studies have shown that steered training data or dataset selection can lead to better performance for cross project defect prediction(CPDP). On the other hand, feature selection and data quality are issues to consider in CPDP. ObjectiveWe aim at utilizing the Nearest Neighbor (NN)-Filter, embedded in genetic algorithm to produce validation sets for generating evolving training datasets to tackle CPDP while accounting for potential noise in defect labels. We also investigate the impact of using different feature sets. MethodWe extend our proposed approach, Genetic Instance Selection (GIS), by incorporating feature selection in its setting. We use 41 releases of 11 multi-version projects to assess the performance GIS in comparison with benchmark CPDP (NN-filter and Naive-CPDP) and within project (Cross-Validation(CV) and Previous Releases(PR)). To assess the impact of feature sets, we use two sets of features, SCM+OO+LOC(all) and CK+LOC(ckloc) as well as iterative info-gain subsetting(IG) for feature selection. ResultsGIS variant with info gain feature selection is significantly better than NN-Filter (all,ckloc,IG) in terms of F1 (p=values≪0.001, Cohen’s d={0.621,0.845,0.762}) and G (p=values≪0.001, Cohen’s d={0.899,1.114,1.056}), and Naive CPDP (all,ckloc,IG) in terms of F1 (p=values≪0.001, Cohen’s d={0.743,0.865,0.789}) and G (p=values≪0.001, Cohen’s d={1.027,1.119,1.050}). Overall, the performance of GIS is comparable to that of within project defect prediction (WPDP) benchmarks, i.e. CV and PR. In terms of multiple comparisons test, all variants of GIS belong to the top ranking group of approaches. ConclusionsWe conclude that datasets obtained from search based approaches combined with feature selection techniques is a promising way to tackle CPDP. Especially, the performance comparison with the within project scenario encourages further investigation of our approach. However, the performance of GIS is based on high recall in the expense of a loss in precision. Using different optimization goals, utilizing other validation datasets and other feature selection techniques are possible future directions to investigate.

Shot boundary detection using perceptual and semantic information

The paper proposes a shot boundary detection system using Gist and local descriptor. Gist provides the perceptual and conceptual information of a scene. The proposed system can be roughly divided into three steps. The first step consists of forming of groups of frames by calculating the correlation of the Gist features between consecutive frames of the video. Secondly, abrupt transitions are found out using the group (G), MSER and a threshold (for abrupt separately, \(th_{cut}\)). And lastly, gradual transitions of the video are found using triangular pattern matching. We have performed the experiment on TRECVid 2001 and 2007 dataset. The novel contribution of this paper is that the proposed system shows an activity-based shot boundary detection where only the possible transition regions of a video are considered for shot detection. This approach reduces the computational complexity by processing the transition regions only. We have achieved better results in terms of F1, precision and recall, when compared to previously published approaches.

A machine‐learning approach to negation and speculation detection for sentiment analysis

Recognizing negative and speculative information is highly relevant for sentiment analysis. This paper presents a machine‐learning approach to automatically detect this kind of information in the review domain. The resulting system works in two steps: in the first pass, negation/speculation cues are identified, and in the second phase the full scope of these cues is determined. The system is trained and evaluated on the Simon Fraser University Review corpus, which is extensively used in opinion mining. The results show how the proposed method outstrips the baseline by as much as roughly 20% in the negation cue detection and around 13% in the scope recognition, both in terms of F1. In speculation, the performance obtained in the cue prediction phase is close to that obtained by a human rater carrying out the same task. In the scope detection, the results are also promising and represent a substantial improvement on the baseline (up by roughly 10%). A detailed error analysis is also provided. The extrinsic evaluation shows that the correct identification of cues and scopes is vital for the task of sentiment analysis.

SUPERVISED TERM WEIGHTING METHODS FOR URL CLASSIFICATION

Many term weighting methods are suggested in the literature for Information Retrieval and Text Categorization. Term weighting method, a part of feature selection process is not yet explored for URL classification problem. We classify a web page using its URL alone without fetching its content and hence URL based classification is faster than other methods. In this study, we investigate the use of term weighting methods for selecting relevant URL features and their impact on the performance of URL classification. We propose a New Relevance Factor (NRF) for the supervised term weighting method to compute the URL weights and perform multiclass classification of URLs using Naive Bayes Classifier. To evaluate the proposed method, we have conducted various experiments on ODP dataset and our experimental results show that the proposed supervised term weighting method based on NRF is suitable for URL classification. We have achieved 11% improvement in terms of Precision over the existing binary classifier methods and 22% improvement in terms of F1 when compared with existing multiclass classifiers.

Analyzing the Cognitive Level of Classroom Questions Using Machine Learning Techniques

Questioning is widely acknowledged as an effective instructional strategy used by teachers in their interaction with students for variety of purposes. In educational practices, the analysis of classroom questions asked by teachers is of particular benefits. This paper investigates the effectiveness of machine learning techniques on analyzing teacher's classroom questions by automatically classifying them into different cognitive levels identified in Bloom's taxonomy. More specifically, this paper reports on three of the most effective machine learning techniques for text classification: k-Nearest Neighbors, Naïve Bayes, and Support Vector Machine using term frequency as a term selection approach. In doing so, a dataset of questions has been collected and classified manually into Bloom's cognitive levels. Preprocessing steps have been applied to convert questions into a representation suitable for machine learning techniques. Using this dataset, the performance of the three machine learning techniques has been evaluated. The results show a comparable performance of k-Nearest Neighbor and Naïve Bayes and a superior performance of Support Vector Machine in term of F1 and accuracy. Moreover the results also indicate that machine learning techniques show different levels of sensitivity to the number of terms used for question representation.

Extracção de Relações Semânticas de Textos em Português Explorando a DBpédia e a Wikipédia

The identification of semantic relationships, as expressed between named entities in text, is an important step for extracting knowledge from large document collections, such as the Web. Previous works have addressed this task for the English language through supervised learning techniques for automatic classification. The current state of the art involves the use of learning methods based on string kernels. However, such approaches require manually annotated training data for each type of semantic relationship, and have scalability problems when tens or hundreds of different types of relationships have to be extracted. This article discusses an approach for distantly supervised relation extraction over texts written in the Portuguese language, which uses an efficient technique for measuring similarity between relation instances, based on minwise hashing and on locality sensitive hashing. In the proposed method, the training examples are automatically collected from Wikipedia, corresponding to sentences that express semantic relationships between pairs of entities extracted from DBPedia. These examples are represented as sets of character quadgrams and other representative elements. The sets are indexed in a data structure that implements the idea of locality-sensitive hashing. To check which semantic relationship is expressed between a given pair of entities referenced in a sentence, the most similar training examples are searched, based on an approximation to the Jaccard coefficient, obtained through min-hashing. The relation class is assigned with basis on the weighted votes of the most similar examples. Tests with a dataset from Wikipedia validate the suitability of the proposed method, showing, for instance, that the method is able to extract 10 different types of semantic relations, 8 of them corresponding to asymmetric relations, with an average score of 55.6%, measured in terms of F1.

Transport Properties of Topological Insulator-Based Ferromagnet/f-Wave Superconductor Junction

We investigate effect of magnetically-induced relativistic mass and also anisotropic f-wave pair coupling on the tunneling conductance on the surface of a 3D topological insulator ferromagnet/superconductor junction, which two types of pairing for superconductivity are possible. A topological insulator as a new state of condensed-matter caused by spin–orbit interaction and time-reversal symmetry has a bulk band gap and gapless surface states. We use the BTK formalism to find the charge carriers behavior. Due to two different nature of order parameters of the f-wave superconductivity, the tunneling conductance is found to be linearly dependent on the magnetic gap in terms of f 1 and the exponential for f 2. It is shown that the conversion of the conductance peak from ZBCP to ZBCD occurs in the f 1 case with increasing m, while this is not observed in f 2. Also, we find that the conductance behaves as a unit step function for the superconductor electrostatic potential in f 1, and this should be usable in nanoelectronic switch devices. In addition, we illustrate how the magnetic gap affects the transmission coefficient in quite different behaviors for order parameters.

RSVD-based Dimensionality Reduction for RecommenderSystems

We investigate dimensionality reduction methods from the perspective of their ability to produce a low-rank customer-product matrix representation. We analyze the results of using collaborative filtering based on SVD, RI, Reflective Random Indexing (RRI) and Randomized Singular Value Decomposition (RSVD) from the perspective of selected algebraic (i.e. application-independent) properties. We show that the Frobenius-norm optimality of SVD does not correspond to the optimal recommendation accuracy, when measured in terms of F1. On the other hand, a high collaborative filtering quality is achievable when a matrix decomposition - based on a combination of RRI and SVD referred to as RSVD-RRI - leads to increased diversity of low-dimensional eigenvectors. The diversity is observable from the perspective of cosine similarities analyzed in comparison to the analogical case of SVD. Such a feature is more desirable than the fidelity of the input matrix spectrum representation, despite the MSE-optimality of SVD.

On Fiber Coefficients of Fiber Cones†

Let (R,𝔪) be a Cohen–Macaulay local ring of dimension d > 0, I an 𝔪-primary ideal of R, and K an ideal containing I. When depth G(I) ≥ d − 1, depth FK(I) ≥ d − 2, and r(I|K) < ∞, we calculate the fiber coefficients fi(I). Under the above assumptions on depth G(I) and r(I|K), we give an upper bound for f1(I), and also provide a characterization, in terms of f1(I), of the condition depth FK(I) ≥ d − 2.

A parametric area function model of three female vocal tracts based on orthogonal modes

Image sets corresponding to 11 vowels were obtained from each of 3 female speakers using Magnetic Resonance Imaging. From these images, vocal tract area functions for each vowel were determined. Each speaker’s 11 area functions were then decomposed into a set of orthogonal basis functions or ‘‘modes’’ and mean vowel area functions. Similar to previous work with male vocal tract data [Story and Titze, J. Phon. (1998)], two orthogonal modes accounted for nearly 90% of the variance within each speaker’s vowel set, while the mean area functions serve as neutral, but unique, tract shapes around which the vowels are formed. For each speaker, the two most significant modes and the mean area function were used to develop a parametric model that allowed for calculation of an arbitrary area function based on coefficients of the modes. With these models a hypothetical vowel space was generated in terms of F1, F2 pairs (first and second formant frequencies). The vowel spaces for each speaker were then used to map time-varying formant frequencies extracted from natural speech to the modal coefficients of their respective parametric models. This effectively produces a time-varying area function suitable for synthesis of speech. [Work supported by NIH R01-DC04789.]

Multiply warped products with nonsmooth metrics

In this article we study manifolds with C0-metrics and properties of Lorentzian multiply warped products. We represent the interior Schwarzschild space–time as a multiply warped product space–time with warping functions and we also investigate the curvature of a multiply warped product with C0-warping functions. We give the Ricci curvature in terms of f1, f2 for the multiply warped products of the form M=(0,2m)×f1R1×f2S2.

MONITORING THROMBIN GENERATION WITH PROTHROMBIN FRAGMENT 1.2 ASSAY DURING CARDIOPULMONARY BYPASS SURGERY

Despite high plasma levels of heparin during cardiopulmonary bypass surgery, activation of the coagulation system has been reported. We hypothesize that the coagulation system activity most appropriately could be assessed by molecular markers of thrombin generation. The aim of the present study was to describe the changes in thrombin generation during CPB, using prothrombin fragment F1+2 (F1.2) as an indicator and evaluate different blood sampling regimens for interpretation of the F1.2 measurements. Twenty patients, operated under extracorporeal circulation with coronary artery bypass grafting (CABG), comprised the study material. The heparin levels were maintained above 2.5 IU /ml throughout the bypass procedure and the functional AT-III level was kept above 0.5 U/ml. Despite of this anticipated inactivation of the coagulation system, the concentrations of F1.2 and FpA increased throughout CPB, particularly after release of the aortic crossclamp. F1.2 and FpA correlated significantly (R=0.69). No statistically significant correlation was found between F1.2 formation rate and age, bodyweight, baseline ACT, ACT after 200 IU heparin/kg, average heparin concentration during CPB or average AT-III level during CPB. CONCLUSIONS: Thrombin formation seems to be a continuous process during CPB despite adequate heparinization. The pattern of thrombin generation can be assessed most appropriately in terms of F1.2 generation rate. Extraordinary high levels of F1.2 were seen after release of the aortic crossclamp, indicating that the periods before and after aortic crossclamping should be evaluated separately.

The generalized damped cubic equation: integrability and general solution

The nonlinear ordinary differential equation of the form u+f1u'+f2u+f3u3=O representing a generalized damped cubic equation is fully analysed. The author finds the general expression for f2 in terms of f1 and f3 that renders the equation integrable. Also he shows that if f2 satisfies the above mentioned condition, the former nonlinear differential equation can always be reduced through adequate changes of variable to the second Painleve transcendent.

Vowel perception in noise

Vowel confusion data presented by Pickett [J. Pickett, J. Acoust. Soc. Am. 29, 613–620 (1957)] were reanalyzed using a multidimensional scaling procedure. The data were analyzed using this method to determine the relative contribution of each dimension under different noise and context conditions. The original data were collected using vowels produced in phonetically balanced words and bVb syllables under different noise spectra conditions. Data were fit with a three‐dimensional solution; two dimensions were comparable to F1 and F2. The subject weights provided estimates of the relative salience of each dimension for each noise condition. Examination of those weights indicated that the relative salience of each dimension varied with vowel context. Dimensions one and three were more heavily weighted for bVb contexts. Dimensions two and three were weighted more heavily for phonetically balanced contexts. Differences in weighting of each dimension yielded perceptual representations that were distortions of the traditional representation of vowel space in terms of F1 and F2.