Historical Behavior Research Articles

Incremental specialized and specialized-generalized matrix factorization models based on adaptive learning rate optimizers

Recommender systems suggest items that are likely to be preferred by a particular user based on historical behavior, actions, and feedback. In real-world applications, data on users and items are continuously generated at a fast pace, such as in e-commerce, social media, digital marketing, and content consumption applications. Since interactions occur over time, these scenarios can be formulated as a data stream where users’ interests are potentially dynamic, i.e., they change over time. Given that changes are expected to occur, one of the current research challenges in streaming recommender systems is that models must adapt their parameters when changes occur to maintain performance. As such changes do not occur for all users and items in the stream at the same time, we consider adapting learning schemes to account for user or item identifiers and model individual parameters. Therefore, we used specialized parameters to adjust the step size for each dataset user or item. More specifically, this study proposes four specialized and specialized-generalized variants of four well-known adaptive learning rate optimizers and shows how they are combined with incremental matrix factorization methods. We tested our proposed optimization strategies on different datasets and showed that one of the proposed specialized variants, that is, InAMSGradUser, improves the RECALL and NDCG rates by up to 11.1 and 7.5 percentage points, respectively, compared to the traditional stochastic gradient descent (SGD) optimizer.

Analysis of Historical Patenting Behavior and Patent Characteristics of Computer Architecture Companies—Part VI: Relationship Between Prosecution Time and Claims

Analysis of Historical Patenting Behavior and Patent Characteristics of Computer Architecture Companies—Part VI: Relationship Between Prosecution Time and Claims

Hydroecological approach for an experimental well field in the Yucatan karst

The experimental well field of FIUADY is a system of 50 m and up to 75 m deep wells. It was built with the purpose of confronting and consolidating the hydrological knowledge of groundwater in the karst plain of northern Yucatan. In this context, the objective of this work is to document the historical behavior of some variables of the physical-chemical and biological environment of the aquifer to sketch models of the habitat of the epikarst microbiota in the northern plain of the Yucatan peninsula. Hydrogeological conformation was documented through drilling debris analysis and underwater video recordings in the well. Measurements of temperature, electrical conductivity, pH and dissolved oxygen were made in October, November, January, February, March and April of different years. For statistical analyses, data were grouped into two periods: recharge (October and November) and discharge (January-April). The results include a schematic hydrogeological profile and statistical analyses of the variables. The minimum and maximum values observed are: temperature °C (26.64, 29.87), electrical conductivity µS/cm (465,46800), pH (6.14, 8.20) and dissolved oxygen mg/L (0.01, 7. 93). It is concluded that: a) in an area of a few square meters there are important variations in the hydrogeological characteristics; b) the hydrogeological behavior of the wells has to do with karstic horizons; their dimensions, connectivity, distribution, density and the way they influence different hydrological stresses; finally c) the presence of crustaceans in the wells and "anomalies" in the concentration of oxygen may have an explanation in biogeochemical processes of dark ecosystems.

Reinforcement Learning Recommendation Algorithm Based on Label Value Distribution

Reinforcement learning is an important machine learning method and has become a hot popular research direction topic at present in recent years. The combination of reinforcement learning and a recommendation system, is a very important application scenario and application, and has always received close attention from researchers in all sectors of society. In this paper, we first propose a feature engineering method based on label distribution learning, which analyzes historical behavior is analyzed and constructs, whereby feature vectors are constructed for users and products via label distribution learning. Then, a recommendation algorithm based on value distribution reinforcement learning is proposed. We first designed the stochastic process of the recommendation process, described the user’s state in the interaction process (by including the information on their explicit state and implicit state), and dynamically generated product recommendations through user feedback. Next, by studying hybrid recommendation strategies, we combined the user’s dynamic and static information to fully utilize their information and achieve high-quality recommendation algorithms. Finally, the algorithm was designed and validated, and various relevant baseline models were compared to demonstrate the effectiveness of the algorithm in this study. With this study, we actually tested the remarkable advantages of relevant design models based on nonlinear expectations compared to other homogeneous individual models. The use of recommendation systems with nonlinear expectations has considerably increased the accuracy, data utilization, robustness, model convergence speed, and stability of the systems. In this study, we incorporated the idea of nonlinear expectations into the design and implementation process of recommendation systems. The main practical value of the improved recommendation model is that its performance is more accurate than that of other recommendation models at the same level of computing power level. Moreover, due to the higher amount of information that the enhanced model contains, it provides theoretical support and the basis for an algorithm that can be used to achieve high-quality recommendation services, and it has many application prospects.

Three-way decision based participants selection optimization model in sparse mobile crowdsensing

Aiming at the problem of low data quality caused by the low quality of participants in sparse mobile crowdsensing, a participant's selection optimization algorithm based on the three-way decision is proposed in this paper. Firstly, a similarity matrix is predicted according to the correlation between Spatio-temporal cells, constructing the classification model. Secondly, the historical behavior of participants is analyzed to extract the characteristic information of the target participants for the task. Finally, with the help of the idea of the three-way decision, the high-quality and reliable candidates are selected according to the participant's characteristic information. Then the participants who perform the task are screened under the guidance of the classification model. The experimental results based on real datasets show that the algorithm proposed in this paper can effectively improve the quality of sensing data and reduce the sensing cost.

A low cost and effective multi-instance abnormal driving behavior detection system under edge computing

The rapid development of intelligent transportation system (ITS) has brought great convenience to people's life, and also makes various intelligent vehicle services emerge in an endless stream. However, with more than 1.35 million people killed in traffic accidents every year, the detection of abnormal driving behavior has attracted more and more attention. Although some progress has been made in the research of abnormal driving behavior detection, numerous challenges remain. Firstly, it is difficult to manually analyze whether driving behavior is abnormal, which makes it costly to accurately label them for machine learning. Secondly, few studies have considered the problem of class imbalance in the abnormal driving behavior detection, which is common in driving behavior and has a significant negative impact on the detection model, especially in weakly supervised scenarios. Thirdly, assessing a driver's behavior solely based on real-time performance is insufficient, the historical behavior is equally important. To address these challenges, an abnormal driving behavior detection system based on weakly supervised learning under edge computing architecture is proposed in this paper. This system considers the class imbalance problem and utilizes the historical information of the vehicles to provide a low cost and effective detection service. Firstly, the abnormal driving behavior detection is modeled as a multi-instance learning (MIL) problem, which only needs coarse-grained labels for anomaly detection, so that the labeling work becomes low cost. Secondly, the class imbalance problem under weakly supervised learning is solved by incorporating parameter transfer and pre-estimating abnormal scores. This approach enhances the model's sensitivity to rare classes and significantly improves the performance of weakly supervised detection models. Thirdly, in order to learn the historical information of the vehicle, a summary mechanism is used for the road side units (RSUs), which can provide different historical summary of the vehicles for other detection nodes, so as to make the model more effective for different abnormal events. Several experiments are carried out on both simulation dataset and real-world dataset to verify the performance of the proposed method. The results demonstrate that compared with the traditional supervised learning methods, the proposed method can still achieve near-optimal performance when only using coarse-grained labels, and all the strategies proposed have improved the detection performance obviously.

Deep Attentive Model for Knowledge Tracing

Knowledge Tracing (KT) is a crucial task in the field of online education, since it aims to predict students' performance on exercises based on their learning history. One typical solution for knowledge tracing is to combine the classic models in educational psychology, such as Item Response Theory (IRT) and Cognitive Diagnosis (CD), with Deep Neural Networks (DNN) technologies. In this solution, a student and related exercises are mapped into feature vectors based on the student's performance at the current time step, however, it does not consider the impact of historical behavior sequences, and the relationships between historical sequences and students. In this paper, we develop DAKTN, a novel model which assimilates the historical sequences to tackle this challenge for better knowledge tracing. To be specific, we apply a pooling layer to incorporate the student behavior sequence in the embedding layer. After that, we further design a local activation unit, which can adaptively calculate the representation vectors by taking the relevance of historical sequences into consideration with respect to candidate student and exercises. Through experimental results on three real-world datasets, DAKTN significantly outperforms state-of-the-art baseline models. We also present the reasonableness of DAKTN by ablation testing.

Use of a Deep Learning Approach for the Evaluation of Students' Online Learning Cognitive Ability

The cognitive ability evaluation of college students' online learning still faces many challenges. Evaluation of college students' online learning cognitive ability is helpful to monitor the learning effect and progress of students, so as to detect problems in time, make suggestions and take corresponding measures to improve. However, the existing evaluation methods for the cognitive ability of college students' online learning often only focus on the degree of knowledge mastery of students, while ignoring the performance of students' learning strategies, autonomous learning ability and emotional attitude in the online learning process. And it is often impossible to understand the learning status of students in real time. To this end, this article studies the evaluation method of college students’ online learning cognitive ability based on deep learning. The evaluation of online learning cognitive ability of college students is divided into two parts: online learning ability evaluation and learning recursive ability evaluation. By introducing the learning ability layer and the learning recursive ability layer into the dynamic key-value memory network and constructing the corresponding evaluation model, it’s better to evaluate the online learning cognitive ability of college students. Based on the historical learning behavior data of students in the online learning process, the online learning cognitive state is predicted, and the performance of the model is further improved through data enhancement and self-attention mechanism. Experimental results verify the validity of the constructed model.

Behavioral factors associated with patients’ non-attendance: A retrospective study in an outpatient specialty clinic at a women’s and children’s hospital in Singapore

ABSTRACT Background This study aimed to identify behavioral factors that affect patient attendance in outpatient clinics, especially those related to rescheduling. Methods A retrospective study was conducted on 20,386 appointment records for 6,252 patients in an outpatient specialty clinic at a women’s and children’s hospital in Singapore over 34 months (January 2012–October 2014). Multivariate logistic regression was used to analyze the influence of independent variables on appointment no-shows. Results The average no-show rate of the study population was 28.87%. Patient historical behaviors were significantly associated with appointment attendance. In particular, a larger number of previous visits, more historical no-shows, more historical rescheduling events initiated by patients, and more reschedulings initiated by the clinic for the current appointment were positively associated with no-shows. Notably, the number of previous visits was found to exhibit a significantly diminishing marginal effect on no-show rates. Further analyses suggested that for appointments rescheduled by the clinic, subsequent rescheduling by patients was associated with a reduced risk of no-shows. Conclusions No-shows were more common in patients who missed more historical appointments or had their appointments rescheduled more frequently. Encouraging patients to reschedule their own appointments could be an effective measure to reduce no-shows. However, the benefit of patient-rescheduling decreased if patients rescheduled appointments repeatedly.

Historic Behavior in Rock–Paper–Scissor Dynamics

Historic Behavior in Rock–Paper–Scissor Dynamics

BERT4Loc: BERT for Location—POI Recommender System

Recommending points of interest (POI) is a challenging task that requires extracting comprehensive location data from location-based social media platforms. To provide effective location-based recommendations, it is important to analyze users’ historical behavior and preferences. In this study, we present a sophisticated location-aware recommendation system that uses Bidirectional Encoder Representations from Transformers (BERT) to offer personalized location-based suggestions. Our model combines location information and user preferences to provide more relevant recommendations compared to models that predict the next POI in a sequence. Based on our experiments conducted on two benchmark datasets, we have observed that our BERT-based model surpasses baselines models in terms of HR by a significant margin of 6% compared to the second-best performing baseline. Furthermore, our model demonstrates a percentage gain of 1–2% in the NDCG compared to second best baseline. These results indicate the superior performance and effectiveness of our BERT-based approach in comparison to other models when evaluating HR and NDCG metrics. Moreover, we see the effectiveness of the proposed model for quality through additional experiments.

Machine learning in aviation: Is it possible to predict when an ATC Sector or Air Traffic Flow will be regulated?

Due to the increasing complexity of airspace, the ATC system does not have sufficient capacity to cope with aircraft demand. For this reason, the ATFCM system needs to implement more and more measures to balance capacity and demand. These measures are the ATFCM regulations. In this paper, a methodology to predict ATFCM capacity regulations based on a machine learning model is proposed. This model will try to predict whether an ATC sector will be regulated or not at a specific time based on the time of prediction and certain operating variables in the sector. A test has been carried out in the LECMPAU sector of Spanish airspace. With results of 91% accuracy in predicting whether the sector will be regulated or not and a logical explainability, it can be concluded that with the sufficient historical operation and certain operational variables, it is possible to predict when an ATFCM regulation capacity will appear in the airspace. It can also be concluded that without a detailed knowledge of the operation in a sector, it is possible to make this prediction because patterns can be found in historical behaviour.

Time-Aware Attention-Based Gated Network for Credit Card Fraud Detection by Extracting Transactional Behaviors

With the popularity of credit cards worldwide, timely and accurate fraud detection has become critically important to ensure the safety of their user accounts. Existing models generally utilize original features or manually aggregated features as their transactional representations, while they fail to reveal the hidden fraudulent behaviors. In this work, we propose a novel model to extract transactional behaviors of users and learn new transactional behavioral representations for credit card fraud detection. Considering the characteristics of transactional behaviors, two time-aware gates are designed in a recurrent neural net unit to learn long- and short-term transactional habits of users, respectively, and to capture behavioral changes of users caused by different time intervals between their consecutive transactions. A time-aware-attention module is proposed and employed to extract the behavioral information from their consecutive historical transactions with time intervals, which enables the proposed model to capture behavioral motive and periodicity inside their historical transactional behaviors. An interaction module is designed to learn more comprehensive and rational representations. To prove the effectiveness of the learned transactional behavioral representations, experiments are conducted on a large real-world transaction dataset and a public one. The results show that the learned representation can well distinguish fraudulent behaviors from legitimate ones, and the proposed method can improve the performance of credit card fraud detection in terms of various evaluation criteria over the state-of-the-art methods.

Construction of College Students’ Management Informatization Ecosystem Based on Data Analysis Technology

College students’ management informatization ecosystem can help college students better manage information and resources, thus improving learning and life efficiency. It’s important to introduce data technology into the college student management informatization ecosystem, which can better realize the extraction of student behavior fluctuation information and achieve more targeted student management decisions. The current research lacks consideration of its future development trend, such as the impact of the application of technologies, including artificial intelligence, big data Fenix, and the Internet of Things. This paper studies the construction of college students’ management informatization ecosystem based on data analysis technology. Firstly, the design of the data fusion scheme applied to the college students’ management informatization ecosystem is carried out, and the idea of the scheme is given. Aiming at the big data of students’ learning and living behaviors, this paper digs deep into the value of data information, and extract historical student behavior characteristics, providing a basis for short-term student behavior prediction and management decision-making planning and implementation for the college students’ management informatization ecosystem. Based on historical student behavior data and its characteristic data, the XGboost algorithm is used to quantify and extract the importance of each influencing factor on student behavior fluctuations. Experimental results verify the effectiveness of the method proposed in this paper.

Application of coordinate systems for vessel trajectory prediction improvement using a recurrent neural networks

According to the Global Maritime Insurance annual report, among human and non-human risk factors, the number of accidents in maritime transport remains a significant issue. One of the factors is vessel collisions and anomalies at sea. Massive historical data from automatic identification systems are analyzed, and intelligent transportation systems are being developed to solve the problem of vessel trajectory prediction. The most ordinary attempt to improve accuracy is by evaluating the historical vessel behavior and learning the patterns and similarities of the predicted vessel movements. However, this paper shows that a better forecast also may be reached by choosing a different trajectory calculation strategy. The geographical or polar coordinate system values are used in a classical way, but several modifications, such as Universal Transverse Mercator (UTM), have been proposed in this study as an alternative to Mercator’s projection coordinates. Two main positioning of the vessels motion transformations were tested: by changing the coordinates to measurements of the angular distance (haversine) and displacement angle (azimuth) functions between different time steps; degree coordinates transformation into a Cartesian system using UTM with vector subtraction. The last case improves the accuracy of almost 30% in the available data sample by using the autoencoder architecture, compared to the longitude and latitude predictions even with computed delta features. The research generally compared three recurrent network architectures (with their hyperparameter – cell sizes): Autoencoder Long Short-Term Memory, Bi-directional Long Short-Term Memory, and Gated Recurrent Unit networks. The model calculations are performed in a real historical dataset, exclusively on cargo vessel type trajectories in the Netherlands (North Sea) coastal region. Also, the methods were validated in another dataset of the Baltic Sea Region.

A firebreak placement model for optimizing biodiversity protection at landscape scale

A solution approach is proposed to optimize the selection of landscape cells for inclusion in firebreaks. It involves linking spatially explicit information on a landscape’s ecological values, historical ignition patterns and fire spread behavior. A firebreak placement optimization model is formulated that captures the tradeoff between the direct loss of biodiversity due to the elimination of vegetation in areas designated for placement of firebreaks and the protection provided by the firebreaks from losses due to future forest fires. The optimal solution generated by the model reduced expected losses from wildfires on a biodiversity combined index due to wildfires by 30% relative to a landscape without any treatment. It also reduced expected losses by 16% compared to a randomly chosen solution. These results suggest that biodiversity loss resulting from the removal of vegetation in areas where firebreaks are placed can be offset by the reduction in biodiversity loss due to the firebreaks’ protective function.

A Double-Layer Blockchain Based Trust Management Model for Secure Internet of Vehicles

The Internet of Vehicles (IoV) enables vehicles to share data that help vehicles perceive the surrounding environment. However, vehicles can spread false information to other IoV nodes; this incorrect information misleads vehicles and causes confusion in traffic, therefore, a vehicular trust model is needed to check the trustworthiness of the message. To eliminate the spread of false information and detect malicious nodes, we propose a double-layer blockchain trust management (DLBTM) mechanism to objectively and accurately evaluate the trustworthiness of vehicle messages. The double-layer blockchain consists of the vehicle blockchain and the RSU blockchain. We also quantify the evaluation behavior of vehicles to show the trust value of the vehicle’s historical behavior. Our DLBTM uses logistic regression to accurately compute the trust value of vehicles, and then predict the probability of vehicles providing satisfactory service to other nodes in the next stage. The simulation results show that our DLBTM can effectively identify malicious nodes, and over time, the system can recognize at least 90% of malicious nodes.

Multi-scale modeling temporal hierarchical attention for sequential recommendation

Multi-scale modeling of items interacting in a sequence of users' historical behaviors in a sequential recommendation task is crucial. In real scenarios, the user's choice of items depends not only on static preferences (long-term interests) but also on recent dynamic preferences (short-term interests). In recent years, micro-video sharing platforms have been very favorable, and correspondingly, more efficient recommendation method will be needed to support users in finding their interested micro-videos (items). Compared to traditional online videos (such as YouTube), micro-videos are created by grassroots users, shot by smartphones, are short (typically tens of seconds), and have fewer tags or descriptive text, which makes recommending it a challenging task. In this work, we explore how to model users' historical behavior sequences at multiple scales to predict their click-through rates on micro-videos and then determine whether to recommend them to users. We present a novel Multi-scale Modeling Temporal Hierarchical Attention (MMTHA) method for modeling users' behavior sequences inspired by recent deep network-based approaches. Specifically, firstly, we capture users' short-term dynamic interests using temporal windows; secondly, we utilize a category-level attention mechanism to describe the coarse-grained interests of users and an item-level attention mechanism to capture the fine-grained interests of users; thirdly, we employ a forward multi-headed self-attention mechanism to identify and integrate long-term correlations between previously segmented temporal windows. We conducted extensive experiments on two publicly available datasets to verify their effectiveness. The experimental results show that our proposed MMTHA model achieves state-of-the-art performance in all tests.

Resilient Supply Chain Framework for Semiconductor Distribution and an Empirical Study of Demand Risk Inference

Supply chain uncertainty is high due to low information transparency in the upstream and downstream, long lead time for supply chain planning, short product life cycles, lengthy production cycle time, and continuous technology migration. The construction and innovation of the new program of supply the chain faces huge challenges. This study aims to propose a smart resilient supply chain framework with a decision-making schema through the plan-do-check-act management cycle. It can enhance supply chain resilience and strengthen industrial competitiveness. Moreover, an empirical study of demand forecast and risk inference for semiconductor distribution is conducted as a validation. Through demand pattern clustering and forecasting for historic customer order behaviors, the demand status of each customer is classified, and an optimal planning solution is released to support decision-making. The result has shown the practical viability of the proposed approach to drive collaborative efforts in enhancing demand risk management to improve supply chain resilience. The proposed forecast model performs better than all four benchmark models, and the revised recall of the proposed risk reference model shows high accuracy in all demand risk levels. As supply chain resilience is about to be reconstructed due to the industrial revolution, a government and industry alliance should follow the resilient supply chain blueprint to gradually make the manufacturing strategy a technology platform in the Industry 4.0 era.

Research on Media resource recommendation System and recommendation Technology based on Graph Neural network

In the field of artificial intelligence research, personalized recommendation technology has been rapidly developed. It can not only provide effective recommendation according to the preferences of system users, but also sort out the historical behavior and attribute information of users, and accurately calculate the content of user preferences. From the perspective of the development of the new media industry, in order to better meet the resource needs of system users, a resource recommendation system based on personalized recommendation technology can fully meet their needs for media reading. Therefore, on the basis of understanding the current research status of media resource recommendation system and technology, and according to the basic concept of graph neural network and the main structure of recommendation system, this paper mainly studies the application effect of the recommendation model of outburst neural network that integrates knowledge graph and time weight, so as to provide effective basis for media resource recommendation in the new era.