Search results

This study aims to construct a sentiment series generation method for danmu comments based on deep learning, and explore the features of sentiment series after clustering.

This study consisted of two main parts: danmu comment sentiment series generation and clustering. In the first part, the authors proposed a sentiment classification model based on BERT fine-tuning to quantify danmu comment sentiment polarity. To smooth the sentiment series, they used methods, such as comprehensive weights. In the second part, the shaped-based distance (SBD)-K-shape method was used to cluster the actual collected data.

The filtered sentiment series or curves of the microfilms on the Bilibili website could be divided into four major categories. There is an apparently stable time interval for the first three types of sentiment curves, while the fourth type of sentiment curve shows a clear trend of fluctuation in general. In addition, it was found that “disputed points” or “highlights” are likely to appear at the beginning and the climax of films, resulting in significant changes in the sentiment curves. The clustering results show a significant difference in user participation, with the second type prevailing over others.

Their sentiment classification model based on BERT fine-tuning outperformed the traditional sentiment lexicon method, which provides a reference for using deep learning as well as transfer learning for danmu comment sentiment analysis. The BERT fine-tuning–SBD-K-shape algorithm can weaken the effect of non-regular noise and temporal phase shift of danmu text.

This study aims to propose a method for monitoring bearing health in the time–frequency domain, termed the Lock-in spectrum, to track the evolution of bearing faults over time and frequency.

The Lock-in spectrum uses vibration signals captured by vibration sensors and uses a lock-in process to analyze specified frequency bands. It calculates the distribution of signal amplitudes around fault characteristic frequencies over short time intervals.

Experimental results demonstrate that the Lock-in spectrum effectively captures the degradation process of bearings from fault inception to complete failure. It provides time-varying information on fault frequencies and amplitudes, enabling early detection of fault growth, even in the initial stages when fault signals are weak. Compared to the benchmark short-time Fourier transform method, the Lock-in spectrum exhibits superior expressive ability, allowing for higher-resolution, long-term monitoring of bearing condition.

The proposed Lock-in spectrum offers a novel approach to bearing health monitoring by capturing the dynamic evolution of fault frequencies over time. It surpasses traditional methods by providing enhanced frequency resolution and early fault detection capabilities.

From these data, the authors construct an uncertainty index through the use of a vector autoregressive (VAR) model to measure the impact of uncertainty on GDP, controlling for inflation, which may affect macroeconomic performance. Results indicate that uncertainty is negatively correlated with the economic cycle and the inter-annual variation of the biannual average product.

This study empirically explores the dynamics of expectations of the Uruguayan manufacturing firms about industrial economic growth. This study explores the dynamics of the industrial economic growth expectations of Uruguayan manufacturing firms. The empirical research is based on firms' expectations data collected through a monthly survey carried out by the Chamber of Industries of Uruguay (CIU) in 2003–2018.

Granger causality tests show that uncertainty Granger-causes industrial production growth and a one standard deviation shock on uncertainty generates a contraction in the industrial production growth rate. Finally, the authors use statistical and network tools to identify groups of firms with similar performance on expectations. Results show that higher uncertainty is associated with smaller, more interconnected groups of firms, and that the number of homogeneous groups and the distance between groups increases with uncertainty. These findings suggest that policies focused on the coordination of expectations can lead to the development of stable opinion groups.

The paper introduces new data and new methodologies to analyze the dynamics of expectations of manufacturing firms about industrial economic growth.

1. An empirical approach to compare expectations of firms is introduced.

2. The occurrence of groups of opinion is tested.

3. Central companies in the network of expectations are detected.

4. More uncertainty implies a higher degree of discrepancy between the overall firm’s opinions and more compact opinion groups.

An empirical approach to compare expectations of firms is introduced.

The occurrence of groups of opinion is tested.

Central companies in the network of expectations are detected.

More uncertainty implies a higher degree of discrepancy between the overall firm’s opinions and more compact opinion groups.

This study aims to introduce an innovative approach to predictive maintenance by integrating time-series sensor data with event logs, leveraging the synergistic potential of deep learning models. The primary goal is to enhance the accuracy of equipment failure predictions, thereby minimizing operational downtime.

The methodology uses a dual-model architecture, combining the patch time series transformer (PatchTST) model for analyzing time-series sensor data and bidirectional encoder representations from transformers for processing textual event log data. Two distinct fusion strategies, namely, early and late fusion, are explored to integrate these data sources effectively. The early fusion approach merges data at the initial stages of processing, while late fusion combines model outputs toward the end. This research conducts thorough experiments using real-world data from wind turbines to validate the approach.

The results demonstrate a significant improvement in fault prediction accuracy, with early fusion strategies outperforming traditional methods by 2.6% to 16.9%. Late fusion strategies, while more stable, underscore the benefit of integrating diverse data types for predictive maintenance. The study provides empirical evidence of the superiority of the fusion-based methodology over singular data source approaches.

This research is distinguished by its novel fusion-based approach to predictive maintenance, marking a departure from conventional single-source data analysis methods. By incorporating both time-series sensor data and textual event logs, the study unveils a comprehensive and effective strategy for fault prediction, paving the way for future advancements in the field.

This study examines the influence of the ownership structure of banks on investors' behavior by dissecting the investors' response to news regarding performance indicators in private and government-owned banks.

The event study methodology is used for the analysis. The data for 35 banks (out of 38), listed on the National Stock Exchange (NSE) for a duration of 230 months (January 2001 to February 2020) is collected. A set of cross-sectional regression analyses is done to identify variables influencing the returns under differential circumstances.

Private banks seem to display a sharper response to negative changes in earnings, while government-owned banks show a more robust reaction to a positive change. The contrast is seen in the variables, having a bearing on the abnormal returns After controlling for a set of factors, the regression analysis shows the ownership structure may not matter on abnormal returns (on event day), the factors such as a change in quarterly earnings, firm-size and three-year average-sales growth influence the positive and negative changes in abnormal returns of government banks, and predictability for private banks is found to be poor regarding selected indicators.

The study evaluates the role of ownership structure on the heterogeneity in investors' responses to the financial performance of banks, thereby assisting in designing strategies to ensure the optimal outcome around the quarterly earnings announcements.

Step length is a key factor for pedestrian dead reckoning (PDR), which affects positioning accuracy and reliability. Traditional methods are difficult to handle step length estimation of dynamic gait, which have larger error and are not adapted to real walking. This paper aims to propose a step length estimation method based on frequency domain feature analysis and gait recognition for PDR, which considers the effects of real-time gait.

The new step length estimation method transformed the acceleration of pedestrians from time domain to frequency domain, and gait characteristics of pedestrians were obtained and matched with different walking speeds.

Many experiments are conducted and compared with Weinberg and Kim models, and the results show that the average errors of the new method were improved by about 2 meters to 5 meters. It also shows that the proposed method has strong stability and device robustness and meets the accuracy requirements of positioning.

A sliding window strategy used in fast Fourier transform is proposed to implement frequency domain analysis of the acceleration, and a fast adaptive gait recognition mechanism is proposed to identify gait of pedestrians.

This paper aims to investigate the trading behavior of insider investors before and after information releases, identifying information-based manipulation in the stock market and the characteristics of companies whose stock prices are manipulated.

This paper employs logit regression method and an event study approach, utilizing hand-collected data from 2010 to 2022, with information categorized into negative and positive types.

The results show no evidence of insider trading or negative information-based manipulation in both high and low transparency firms. However, in highly transparent companies, the Board of Directors (BOD) avoids direct manipulation by using relatives to evade market supervisors. In low transparency companies, both the BOD and family members (FM) exploit positive information to benefit personally by buying shares before releasing favorable news, causing a sharp stock increase, and selling afterward. Continued buying by the BOD and FM also suggests likely positive news announcements.

The characteristics of information-based manipulation in companies, as provided by this study, help individual investors avoid investing in stocks that are highly susceptible to manipulation.

Empirical research on information-based manipulation is scarce due to limited secondary data. Our study uses transaction data from insider investors in a frontier market with low transparency and high information asymmetry. This enables us to analyze information-based stock price manipulation. We identify manipulation by comparing insiders' trading behavior with their market information releases, resulting in stock price fluctuations greater than 5%.

The role of inventors' creativity is crucial for technological innovation within enterprises. The mobility of inventors among different enterprises is a primary source for companies to acquire external knowledge. The mechanism of “learning-by-hiring” is widely recognized by companies. Therefore, it is important to determine how to allocate network resources to enhance the creativity of inventors when companies hire mobile inventors.

The study suggests an analytical framework that analyzes alterations in tie strength and structural holes resulting from the network embeddedness of mobile inventors as well as the effect of the interaction between these two variables on changes in inventor’s creativity after the mobility. In addition, this paper examines the moderating impact of cognitive richness of mobile inventors and cognitive distance between mobile inventors and new employers on the correlation between network embeddedness and creativity.

This study found that: (1) The increase of tie strength has a significant boost in creativity. (2) Increasing structural holes can significantly improve the creativity of mobile inventors. (3) When both the tie strength and the structural holes increase, the creativity of the mobile inventors significantly increases. (4) It is important to note that when there is a greater cognitive distance, stronger tie strength promotes the creativity of mobile inventors. Additionally, cognitive richness plays a significant role in moderating the relationship between changes in structural holes and the creativity of mobile inventors.

These findings provide theoretical guidance for firms to effectively manage mobile inventors and optimize collaborative networks within organizations.

The criticality of late deliveries in transportation lies in the threat of considerable multi-level supply chain costs. This study aims to reveal the dynamic capabilities playing a facilitating role in preventing delay, thus providing timely delivery, as well as developing an understanding of how and when those capabilities are activated within the supply chain network.

An exploratory study was conducted involving 16 semi-structured expert interviews with the representatives of logistics service providers and shippers. Following an interpretive phenomenology framework, the prevention phenomenon was explained.

Findings revealed two preventive capability categories in delay prevention: (1) proactive capabilities, referring to the enabling actions planned before departure, and (2) reactive capabilities, referring to actions planned after departure. Findings pinpoint that, in addition to the proactive capabilities, reactive capabilities enabled by innovative problem-solving actions are crucial for adapting to a dynamically changing environment in prevention. Moreover, this study shows that prevention capabilities are characterized by tangible and intangible resources and integration of resources with external links which constitute a delay prevention network within a wider service ecosystem.

This study stands out with its specific focus on delay prevention capabilities and enabling actions from the perspectives of logistics service providers and shippers. The premises of the resource-based view are combined with dynamic capabilities theory, leading to a proposed time-based taxonomy of proactive and reactive capabilities in supply chains, aimed at creating value and strengthening resilience.

The accurate prediction of driving torque demand is essential for the development of motion controllers for mobile robots on complex terrains. This paper aims to propose a hybrid model of torque prediction, adaptive EC-GPR, for mobile robots to address the problem of estimating the required driving torque with unknown terrain disturbances.

An error compensation (EC) framework is used, and the preliminary prediction driving torque value is achieved using Gaussian process regression (GPR). The error is predicted using a continuous hidden Markov model to generate compensation for the prediction residual caused by terrain disturbances and uncertainties. As the final step, a gain coefficient is used to adaptively tune the significance of the compensation term through parameter resetting. The proposed model is verified on a sample set, including the driving torque of a mobile robot on three different sandy terrains with two driving modes.

The results show that the adaptive EC-GPR yields the highest prediction accuracy when compared with existing methods.

It is demonstrated that the proposed model can predict the driving torque accurately for mobile robots in an unconstructed environment without terrain identification.