Accurate Similarity Measure Research Articles

Measuring the similarity between shapes of buildings using graph edit distance

ABSTRACT Building data play a crucial role in geographic information science, and building shape similarity can be used for spatial inquiry, cartographic generalization, map updating, data quality assessment, and other spatial data mining applications. However, conventional shape representation and measurement methods use geometric and statistical measures and may not effectively capture the subtle differences among building shapes. Geospatial artificial intelligence methods often overlook nuanced differences in building contours, posing a challenge when precise fine-grained measurements are required. Furthermore, they have low interpretability. This study proposes a graph edit distance-based method for measuring the similarity between building shapes. It uses geometrics characteristics to construct a graph, matches and associates similar regional features between two building contours, and defines edit costs based on geometric and positional differences that align with the Gestalt principle of visual similarity. It converts the optimal edit path problem to the maximum weighted cluster solution problem and utilizes a heuristic approach to achieve the exact optimal solution within a moderately suitable timeframe. The experimental results demonstrated that the proposed method identifies detailed differences between building contours, providing accurate similarity measurement with high interpretability.

The Bi–O edge wavefront sensor

Context. Direct detection of exoplanets around nearby stars requires advanced adaptive optics (AO) systems. High-order systems are needed to reach a high Strehl ratio (SR) in near-infrared and optical wavelengths on future giant segmented-mirror telescopes (GSMTs). Direct detection of faint exoplanets with the European Southern Observatory (ESO) Extremely Large Telescope (ELT) will require some tens of thousands of correction modes. The resolution and sensitivity of the wavefront sensor (WFS) are key requirements for this science case. We present a new class of WFSs, the bi-orthogonal Foucault knife-edge sensors (or Bi–O edge), that is directly inspired by the Foucault knife-edge test. The idea consists of using a beam-splitter producing two foci, each of which is sensed by an edge with a direction orthogonal to the other focus. Aims. We describe two implementation concepts: The Bi–O edge sensor can be realised with a sharp edge and a tip-tilt modulation device (sharp Bi–O edge) or with a smooth gradual transmission over a grey edge (grey Bi–O edge). A comparison of the Bi–O edge concepts and the four-sided classical pyramid wavefront sensor (PWS) gives some important insights into the nature of the measurements. Methods. We analytically computed the photon noise error propagation, and we compared the results to end-to-end simulations of a closed-loop AO system. Results. Our analysis shows that the sensitivity gain of the Bi–O edge with respect to the PWS depends on the system configuration. The gain is a function of the number of control modes and the modulation angle. We found that for the sharp Bi–O edge, the gain in reduction of propagated photon noise variance approaches a theoretical factor of 2 for a large number of control modes and small modulation angle, meaning that the sharp Bi–O edge only needs half of the photons of the PWS to reach similar measurement accuracy. In contrast, the PWS is twice more sensitive than the Bi–O edge in the case of very low order correction and/or large modulation angles. Preliminary end-to-end simulations illustrate some of the results. The grey version of the Bi–O edge opens the door to advanced amplitude filtering, which replaces the need for a tip-tilt modulator while keeping the same dynamic range. We show that an additional factor of 2 in reduction of propagated photon noise variance can be obtained for high orders, such that the theoretical maximum gain of a factor of 4 in photon efficiency can be obtained. A diffractive Fourier model that accurately includes the effect of modulation and control modes shows that for the extreme AO (XAO) system configuration of the ELT, the overall gain will well exceed one magnitude in guide-star brightness when compared to the modulated PWS. Conclusions. We conclude that the Bi–O edge is an excellent candidate sensor for future very high order Adaptive Optics systems, in particular on GSMTs.

Can three-dimensional models enhance understanding and knowledge of rotator cuff tears?

ObjectivesMagnetic resonance imaging (MRI) is currently the standard diagnostic tool for rotator cuff tears. However, its two-dimensional (2D) output, displayed on a monitor, can complicate the interpretation of anatomy. Three-dimensional (3D) imaging may offer a solution to this issue. This study aimed to demonstrate the diagnostic and interpretive value of a 3D model in assessing lesion anatomy. The hypothesis was that 3D models, compared to 2D MRI, can enhance the comprehension and knowledge of rotator cuff injuries, improve the application of classifications for total tears, and provide a more precise definition of the size and type of tear. MethodsA prospective single-centre study was conducted. 3D models for rotator cuff tears were created and analysed in conjunction with preoperative MRI for each patient up to 2 months before surgery. The 3D models were based on the preoperative MRI. Collected data included 2D plane measurements by MRI in coronal and sagittal planes, descriptions of 3D lesion geometry (new shapes), 3D measurements in coronal and sagittal planes, arthroscopic classifications of rotator cuff injuries, and arthroscopic measurements in coronal and sagittal planes. ResultsAfter examining 25 cases, 3D imaging demonstrated similar arthroscopic values post-bursectomy in the sagittal plane (16.70 ​mm for 3D and 18.28 ​mm for post-bursectomy, p-value ​= ​0.189), although these measurements did not align with those of MRI (which underestimated measurements, p-value ​= ​0.010). Both MRI measurement and 3D imaging showed similar measurement accuracy in the coronal plane when compared to arthroscopic measurements taken before and after bursectomy. The creation of 3D objects enabled the analysis of new geometries, including the length, width, and depth of each lesion. These geometries included the rectangle, rectangular trapezoid, scalene trapezoid, irregular pentagon, and irregular hexagon. Conclusions3D models can enhance the understanding and knowledge of rotator cuff injuries. They can be a promising tool for diagnosing and interpreting the anatomy of the injury, particularly in the sagittal plane. The new 3D understanding of the pathological process has led to the description of new geometric features not visible in conventional 2D MRI. Level of evidenceII – Development of diagnostic criteria on consecutive patients (all compared to “gold” standard).

Adaptive batch mode active learning with deep similarity

Active learning is usually used in scenarios where few labels are available and manual labeling is expensive. To improve model performance, it is necessary to find the most valuable instance among all instances and label it to maximize the benefits of labeling. In practical scenarios, it is often more efficient to query a group of instances instead of a individual instance during each iteration. To achieve this goal, we need to explore the similarities between instances to ensure the informativeness and diversity. Many ad-hoc algorithms are proposed for batch mode active learning, and there are generally two major issues. One is that similarity measurement among instances often only relies on the expression of features but it is not well integrated with the classification algorithm model. This will cut down the precise measurement of diversity. The other is that in order to explore the decision boundary, these algorithms often choose the instance near the boundary. It is difficult to get the true boundary when there are few labeled instances. As a large number of instances continue to be labeled, information between instances is less used, and the performance will be greatly improved if it is properly used. In our work, we propose an adaptive algorithm based on deep neural networks to solve the two problems mentioned above. During the training phase, we established a paired network to improve the accuracy of the classification model, and the network can project the instance to a new feature space for more accurate similarity measurement. When batch labeling instances, we use the adaptive algorithm to select the instance by balancing the maximum uncertainty (exploration) and diversity (exploitation). Our algorithm has been validated for heart failure prediction tasks in real-world EHR datasets. Due to the no public of EHR data, we also conducted validation on two other classic classification tasks. Our algorithm is superior to the baseline method in both accuracy and convergence rate.

An Efficient and Accurate Convolution-Based Similarity Measure for Uncertain Trajectories

With the rapid development of localization techniques and the prevalence of mobile devices, massive amounts of trajectory data have been generated, playing essential roles in areas of user analytics, smart transportation, and public safety. Measuring trajectory similarity is one of the fundamental tasks in trajectory analytics. Although considerable research has been conducted on trajectory similarity, the majority of existing approaches measure the similarity between two trajectories by calculating the distance between aligned locations, leading to challenges related to uncertain trajectories (e.g., low and heterogeneous data sampling rates, as well as location noise). To address these challenges, we propose Contra, a convolution-based similarity measure designed specifically for uncertain trajectories. The main focus of Contra is to identify the similarity of trajectory shapes while disregarding the time/order relevance of each record within the trajectory. To this end, it leverages a series of convolution and pooling operations to extract high-level geo-information from trajectories, and subsequently compares their similarities based on these extracted features. Moreover, we introduce efficient trajectory index strategies to enhance the computational efficiency of our proposed measure. We conduct comprehensive experiments on two trajectory datasets to evaluate the performance of our proposed approach. The experiments on both datasets show the effectiveness and efficiency of our approach. Specifically, the mean rank of Contra is 3 times better than the state-of-the-art approaches, and the precision of Contra surpasses baseline approaches by 20–40%.

Explainable Image Similarity: Integrating Siamese Networks and Grad-CAM.

With the proliferation of image-based applications in various domains, the need for accurate and interpretable image similarity measures has become increasingly critical. Existing image similarity models often lack transparency, making it challenging to understand the reasons why two images are considered similar. In this paper, we propose the concept of explainable image similarity, where the goal is the development of an approach, which is capable of providing similarity scores along with visual factual and counterfactual explanations. Along this line, we present a new framework, which integrates Siamese Networks and Grad-CAM for providing explainable image similarity and discuss the potential benefits and challenges of adopting this approach. In addition, we provide a comprehensive discussion about factual and counterfactual explanations provided by the proposed framework for assisting decision making. The proposed approach has the potential to enhance the interpretability, trustworthiness and user acceptance of image-based systems in real-world image similarity applications.

Local morphological patterns for time series classification

The key problem of time series classification is the similarity measure between time series. In recent years, efficient and accurate similarity measurement methods of time series have attracted extensive attention from researchers. According to the different similarity measure strategies, the existing time series classification methods can be roughly divided into shape-based (original value) methods and structure-based (symbol transformation) methods. Shape-based methods usually use Euclidean distance (ED), dynamic time warping (DTW), or other methods to measure the global similarity between sequences. The disadvantage of these methods is that their measurement process does not necessarily achieve local sensible matchings of time series, which leads to a decrease in their accuracy and interpretability. To better capture the local information of the sequence, the structure-based methods discretize or symbolize the local value of the time sequence, which leads to the loss of the original information of the sequence. To address these problems, this paper proposes a novel similarity measurement method named dynamic time warping based on the local morphological pattern (MPDTW), which first decomposes the local subsequences of time series using discrete wavelet transforms for extracting the local structure information. Then, the decomposed subsequence will be encoded by the morphological pattern. Finally, the ED between points and their local structure difference based on morphological pattern will be weighted and applied to the DTW algorithm to measure the similarity between sequences. Experiments have been carried out on the classification tasks of the UCR datasets and the results show that our method outperforms the existing baselines.

ComparePD: Improving protein-DNA complex model comparison with hydrogen bond energy-based metrics.

Computational modeling of protein-DNA complex structures has important implications in biomedical applications such as structure-based, computer aided drug design. A key step in developing methods for accurate modeling of protein-DNA complexes is similarity assessment between models and their reference complex structures. Existing methods primarily rely on distance-based metrics and generally do not consider important functional features of the complexes, such as interface hydrogen bonds that are critical to specific protein-DNA interactions. Here, we present a new scoring function, ComparePD, which takes interface hydrogen bond energy and strength into account besides the distance-based metrics for accurate similarity measure of protein-DNA complexes. ComparePD was tested on two datasets of computational models of protein-DNA complexes generated using docking (classified as easy, intermediate, and difficult cases) and homology modeling methods. The results were compared with PDDockQ, a modified version of DockQ tailored for protein-DNA complexes, as well as the metrics employed by the community-wide experiment CAPRI (Critical Assessment of PRedicted Interactions). We demonstrated that ComparePD provides an improved similarity measure over PDDockQ and the CAPRI classification method by considering both conformational similarity and functional importance of the complex interface. ComparePD identified more meaningful models as compared to PDDockQ for all the cases having different top models between ComparePD and PDDockQ except for one intermediate docking case.

A case weighted similarity deep measurement method based on a self-attention Siamese neural network

To improve the accuracy of similarity measures in case-based reasoning, in this paper, we propose a deep metric learning method based on a self-attention mechanism and a Siamese neural network to realize the weighted similarity measure between cases. In this method, weight assignment and similarity measurement processes are integrated into a deep network. The method can map cases to a new feature space through nonlinear processing of the network layer to obtain better feature representation. The inner relationship between the features is captured by the self-attention mechanism, which is connected to the previous network layer, and the weight of the features is determined by the scoring function. Finally, a metric function is added to the contrastive loss to measure the case similarity. Experiments show that the accuracy of this method is better than that of other algorithms in the similarity measure and can improve the accuracy of case retrieval.

Accuracy improvement of surface measurement through phase correction in spectrally resolved interferometer

The spectrally resolved interferometer (SRI) is used to measure the surface profile by the linear fitting line of its phase distribution of spectral interference signal. However, the noise existed in the phase cannot be ignored if the nano-scale measurement accuracy is required. In this paper, a novel algorithm is proposed to improve the measurement accuracy by using the constant term oflinear fitting line of phase distribution in wavenumber domain. The simulations indicate that the algorithm achieves the similar measurement accuracy under the influence of three kinds of noise. Even though the proposed algorithm has the possibility of 4% to bring the additional minor error, it provides a better performance on measurement of entire surface in different SNRs compared with the other two common algorithms. In experiments, the variations of surface less than 10 nm and 30 nm and the repeatability of 2.2 nm and 4.9 nm are obtained in the measurements of a protected aluminium mirror with and without proposed algorithm. A smooth profile with a step shape of nearly 3 μm height is tested, where the maximum repeatability of step height and maximum magnitude of variations of the surfaces reaches to 3.8 nm and 10 nm from the 15.3 nm and 30 nm by using this algorithm. Both the simulations and experimental results demonstrate that the proposed algorithm is feasible for high-precision measurement.

Evaluation of Low-Cost UV-Vis Spectroscopy for Measuring Nitrate Using Synthetic Water Samples

Highlights A novel low-cost, modular spectroscopy system is compared to a standard system using synthetic samples, principal component analysis, and partial least squares regression. The information contained in the data produced by the two systems is similar according to principal component analysis. The low-cost system was able to accurately predict nitrate concentrations in concentrated and diluted samples using partial least squares regression. The methodology could be applied to water quality analysis in agriculture and water resources management. Abstract. Water quality data collection is an essential component of water systems management. For instance, the effective management of nutrients in hydroponic systems is necessary for maximizing yields efficiently and sustainably. Additionally, nutrients in natural and engineered waterbodies must be monitored to ensure they are meeting the required chemical characteristics for their ecological and social functions. However, conventional water quality data collection methods place limitations on water systems management due to their high resource requirements. Nitrate (NO3) is a major nutrient in ecological and agricultural systems, which can be reliably measured with ultraviolet-visible (UV-Vis) spectroscopy, a highly established technique for water quality analysis. The goal of this research was to evaluate a novel, low-cost, modular UV-Vis spectroscopy setup (GatorSpec) for the measurement of NO3 concentration in chemically complex solutions. UV-Vis absorbance of synthetic samples was measured using the GatorSpec and a commonly used bench-top laboratory spectroscopy system, the NanoDrop2000C. These data were analyzed using principal component analysis (PCA) to compare the spectral data produced by each system and partial least squares (PLS) regression to compare their ability to predict NO3 concentration. Results showed that data from both measurement systems were similar, indicating that the low-cost GatorSpec provided similar measurement accuracy to that of the laboratory reference system, the NanoDrop2000C. The PLS results revealed that for the diluted samples, the models derived from both systems were very good at predicting NO3 concentration. With these outcomes, it can be concluded that the GatorSpec is effective at measuring NO3 concentration in complex solutions and is comparable in performance to that of the NanoDrop2000C. In the future, this low-cost setup could be used to manage NO3 concentrations more efficiently in various applications such as hydroponic plant production, environmental monitoring, and stormwater treatment, which, in turn, could reduce the economic and environmental costs of these systems. Keywords: Low-cost, Synthetic samples, Ultraviolet-visible absorption spectroscopy, Water quality.

Three-Dimensional Measurement Method for High Dynamic Range Surfaces Based on Adaptive Fringe Projection

As an optical measurement method, fringe projection three-dimensional (3D) measurement technology will have poor measurement effect due to overexposure or too dark of the fringes when measuring high dynamic range (HDR) surfaces. The adaptive fringe projection method is one of the effective methods to solve the HDR surfaces measurement, but pixel matching is the difficulty. The existing matching methods generally use phase for matching, which requires at least 24 additional patterns to be projected, seriously affecting the measurement efficiency. In order to solve these problems, an adaptive fringe projection measurement method using speckle pattern for pixel matching is proposed in this paper. First, two grayscale patterns are projected to calculate the adaptive projection intensity. Then a speckle pattern is projected to match the pixels of the captured pattern and the projected pattern. Finally, the adaptive fringe patterns are generated. Better than the methods before, only three additional patterns are needed to be projected to realize the measurement of the HDR surfaces. The experimental results show that the proposed method can reduce the number of projected patterns by more than 59.5% while achieving similar measurement accuracy, which greatly improves the measurement efficiency.

Strengthened multiple correlation for multi-label few-shot intent detection

A challenging problem that arises in few-shot intent detection is the complexity of multiple intention (multi-label) detection. The prototypical network uses the mean value of support instances as label prototype, which cannot eliminate the interference among features of multiple labels, making the learned label prototypes deviate from the real label features. Meanwhile, regardless of the correlation with the label prototype, all the feature dimensions in the query instance are treated equally, which reduces the accuracy of similarity measurement between the query instance and the label prototype. In this paper, we propose a hybrid calculation of correlation for few-shot multi-label intent detection method (HCC-FSML) to overcome the problem. This method proposes an instance-level attention mechanism to focus on the instance representation with high correlation between support instances and positive labels, so as to improve the consistency between the label prototype and the real label features; In the similarity measurement, the feature-level attention mechanism is introduced to focus on the feature dimensions of query instance with high correlation with positive label prototype, so as to improve the accuracy of similarity measurement. The experiment shows that the proposed method achieves new state-of-the-art results in intent detection by focusing on typical features of instances and reducing multiple interference among labels.

A Low-Cost Wireless Sensor for Real-Time Monitoring of Water Level in Lowland Rice Field under Alternate Wetting and Drying Irrigation

The use of wireless sensors for real-time monitoring of field water level would greatly facilitate the application of alternate wetting and drying (AWD), an irrigation water management technique proven to result to significant water savings and reduced methane emissions in lowland rice production systems. However, most of the commercially available wireless sensors are generally costly. This study developed a low-cost wireless sensor that can perform real-time monitoring of water depth and surface temperature in lowland rice fields under an AWD irrigation regime. The sensor is composed mainly of an ultrasonic depth sensor, a waterproof temperature sensor, a humidity sensor, and a Wi-Fi-enabled microcontroller enclosed in a PVC cap that can be mounted in AWD pipes. The sensor was tested under laboratory, pseudo-field conditions and actual field conditions. Results showed a relatively high degree of agreement between sensor and manual measurements of water depth under all testing conditions, with the error ranging from only 5.2% to 6.6% and RMSE of 5.0 mm to 13.5 mm. The performance of the low-cost sensor also proved to be comparable with that of the high-end sensor, exhibiting practically similar measurement accuracy and higher precision. The wireless sensor developed in this study can provide a low-cost alternative to the high-cost and high-end sensors and other commercially available counterparts for efficient irrigation water management in lowland crop production systems during water-scarce conditions induced by climate change and climate variability.

Two efficient selection methods for high-dimensional CD-CAT utilizing max-marginals factor from MAP query and ensemble learning approach.

Computerized adaptive testing for cognitive diagnosis (CD-CAT) needs to be efficient and responsive in real time to meet practical applications' requirements. For high-dimensional data, the number of categories to be recognized in a test grows exponentially as the number of attributes increases, which can easily cause system reaction time to be too long such that it adversely affects the examinees and thus seriously impacts the measurement efficiency. More importantly, the long-time CPU operations and memory usage of item selection in CD-CAT due to intensive computation are impractical and cannot wholly meet practice needs. This paper proposed two new efficient selection strategies (HIA and CEL) for high-dimensional CD-CAT to address this issue by incorporating the max-marginals from the maximum a posteriori query and integrating the ensemble learning approach into the previous efficient selection methods, respectively. The performance of the proposed selection method was compared with the conventional selection method using simulated and real item pools. The results showed that the proposed methods could significantly improve the measurement efficiency with about 1/2-1/200 of the conventional methods' computation time while retaining similar measurement accuracy. With increasing number of attributes and size of the item pool, the computation time advantage of the proposed methods becomes more significant.

Measurement Model for Medical Image Feature Matrix Similarity Based on CNN

The original similarity measurement model is easy to ignore the processing of image details, resulting in poor accuracy of similarity measurement. In the paper, we propose a similarity measurement model for the medical image feature matrix based on the convolutional neural network (CNN). First, the Gaussian convolution kernel is used to obtain the global and local feature data of medical images, and the corresponding data set is formed. Second, the convolution layer of CNN is introduced, and the image feature matrix is obtained by the convolution layer. Finally, the similarity measurement model of the medical image feature matrix is constructed. The results show that the image similarity measurement effect of this model is better when the test process is divided into three parts: global, local, and detail. The highest error rate of the proposed algorithm is only about 0.21, which takes less time, and the overall fitting degree can reach about 91%. Compared with traditional methods, the accuracy of image similarity measurement is higher and the use effect is better.

Seeded Classification of Satellite Image Time Series with Lower-Bounded Dynamic Time Warping

Satellite Image Time Series (SITS) record the continuous temporal behavior of land cover types and thus provide a new perspective for finer-grained land cover classification compared with the usual spectral and spatial information contained in a static image. In addition, SITS data is becoming more accessible in recent years due to newly launched satellites and accumulated historical data. However, the lack of labeled training samples limits the exploration of SITS data, especially with sophisticated methods. Even with a straightforward classifier, such as k-nearest neighbor, the accuracy and efficiency of the SITS similarity measure is also a pending problem. In this paper, we propose SKNN-LB-DTW, a seeded SITS classification method based on lower-bounded Dynamic Time Warping (DTW). The word “seeded” indicates that only a few labeled samples are required, and this is not only because of the lack of labeled samples but also because of our aim to explore the rich information contained in SITS, rather than letting training samples dominate the classification results. We use a combination of cascading lower bounds and early abandoning of DTW as an accurate yet efficient similarity measure for large scale tasks. The experimental results on two real SITS datasets demonstrate the utility of the proposed SKNN-LB-DTW, which could become an effective solution for SITS classification when the amount of unlabeled SITS data far exceeds the labeled data.

Case-based reasoning system for fault diagnosis of aero-engines

Fault diagnosis in aero-engines typically requires an experienced expert for understanding and detecting the cause of faults. However, accurate and quick identification of fault parts is difficult for maintenance crews owing to the complexity of aero-engines. In this study, we developed a case-based reasoning (CBR) system with a highly accurate novel similarity measure for fault diagnosis of aero-engines by retrieving similar fault cases. The proposed CBR system is established based on 143 cases with the knowledge of correctly diagnosed and successfully resolved aero-engine faults, which constitutes the first tentative case base in the field of aero-engine fault diagnosis. The proposed case similarity measure for fault diagnosis of aero-engines integrates three local similarity measures associated with different attributes, especially among which a tree-based semantic similarity measure is proposed to define the relationship between the fault part and fault mode based on a semantic graph incorporated into the aero-engine tree structure. The proposed CBR system is evaluated using the k-nearest neighbors algorithm and 5-fold cross-validation. The system exhibited high retrieval accuracy with all cases collected from real-world scenarios of aero-engine fault diagnosis. Our study shows great promise that the experience-based decisions yielded from the results can aid in aero-engine maintenance and support services.

Weight coefficient calculation through equivalent ray tracing method for light field particle image velocimetry

Light field particle image velocimetry (LF-PIV) can measure three-dimensional (3D) flow velocity from a single snapshot of a light field camera based on the 3D reconstruction of tracer particles of light field images. However, it requires light field intensity calibrations to calculate accurate weight coefficients. Conventionally, the weight coefficients are calculated through in-situ calibration approaches whereas the translation of the calibration board within the entire measurement area is required. Therefore, these approaches are inapplicable for internal industrial flows and space-constrained applications. This study presents an equivalent ray tracing method for the weight coefficients calculation. With a light field snapshot of a smart calibration board, a mapping relationship is established to relate the target points sampled in the flow field with their equivalent points in the air. The weight coefficients are then calculated through the ray tracing method by changing the starting points of ray tracing from the target points to their equivalent points. The calculated weight coefficients are employed to reconstruct the feature marks of the calibration board. Results show that the spatial locations of the marks can be reconstructed accurately with a mean lateral and depth error of 0.63% and 5.6%, respectively. Experiments were also carried out on a low-speed laminar flow. The result indicates that the equivalent ray tracing method provides a similar measurement accuracy with the in-situ calibration method. The overall error of 6.77% is achieved for the velocity measurement. It is demonstrated that the proposed method is capable of measuring the 3D flow velocity for internal industrial and space-constrained applications without translating the calibration board within the entire measurement area.

Measuring Similarity Between Any Pair of Passengers Using Smart Card Usage Data

Recent years have witnessed considerable progress in the application of Internet of Things (IoT) technology in smart transportation systems. The wider presence of Wi-Fi networks in subway gates allows passengers to use the quick response (QR) code of mobile phone applications for entrance. The network established by gates has become a medium which connects stations and passengers. However, in addition to directly monitoring the passenger flow, the potential application of the smart card usage data collected by the gates remains an open topic. Although there are several clustering-based works devoted to revealing passengers’ travel behavior patterns, research on the social attributes of subway passengers is very limited. To fill the gap, this article proposes a novel method to mine similarity information of passengers by leveraging passengers’ communication behaviors hidden in subway card usage data. Passengers are first organized as a graph, which not only reflects the interactions between them but also incorporates the context information of subway stations. Then, the node embedding is used to encode the information contained in the graph and with the use of cosine similarity, the similarity between two passengers is measured. Extensive experiments on two real-world location-based social network data sets and extended experiments on a Shanghai subway data set are conducted. The results show that the proposed method can effectively improve the accuracy of similarity measurement and provide social features that are distinguishable from travel behavior patterns.