China
Africa
Explore chapters and articles related to this topic
Case Studies
Published in Abhijit Pandit, Mathematical Modeling using Fuzzy Logic, 2021
The vector autoregression (VAR) method uses an AR model to model the next step in each time series. It is a generalization of AR to multiple parallel time series. Multivariate time series: Model notation involves specifying the order of the AR(p) model as a parameter to the VAR function, e.g. VAR(p). This method is suitable for multivariate time series without trend and seasonal components.
State-of-the-Art Research in the Area of Artificial Intelligence with Specific Consideration to Civil Infrastructure, Construction Engineering and Management, and Safety
Published in M.Z. Naser, Leveraging Artificial Intelligence in Engineering, Management, and Safety of Infrastructure, 2023
Islam H. El-Adaway, Rayan H. Assaad
The automatic recognition of labor actions was possible due to the exceptional capabilities of ML techniques (namely k-nearest neighbors, multilayer perceptron, decision trees, and support vector machines) (Ryu et al., 2019). K-nearest neighbors were also used with case-based reasoning for the planning of deep foundation construction (Zhang et al., 2017). Studying the impact of dynamic workforce and workplace variables on the productivity of the construction industry was performed by relying on time series econometric analysis based on vector autoregression models (Assaad and El-adaway, 2021b), Productivity Indicator. In addition, association rule mining and Bayesian networks were employed to identify the relationships between defects and their occurrence probabilities (Fan, 2020). Furthermore, a method for detecting concrete structural components in color images was proposed based on image processing, Gaussian mixture model, artificial neural networks, and support vector machines (Son et al., 2012). Decision trees and naïve Bayesian classification were used for facilitating accurate project delay risk analysis and prediction (Gondia et al., 2020) to help in addressing schedule overruns that are highly present in the construction industry (Assaad et al., 2020c). Moreover, the integration of field submittals in project scheduling has been performed using random forests by mining project data to forecast delay during the project and to forecast the likelihood of acceptance of submittal requests (Awada et al., 2021). Additionally, clustering (expectation maximization) was utilized to propose an approach that can gather valuable knowledge from previously unanalyzed data to significantly improve resource and labor management practices (Hammad et al., 2014). Reinforcement learning was used to examine the impact of learning in bidding decision-making processes to help in having optimal outcomes in the long run (Assaad et al., 2021). Another application of AI-techniques in construction- related activities is the estimation of schedule to completion through neural network– long short-term memory (feedforward neural networks for nonsequential issue and recurrent neural networks for sequential issue) (Cheng et al., 2019). Long short-term memory architectures were also used with natural language processing to develop an automated specification reviewing model and an information extraction framework that is used for construction specification review (Moon et al., 2021). Chow–Liu tree and k-means clustering were used to develop a framework that measures the amount of uncertainty and information sharing among constraint discussions in construction planning meetings (Javanmardi et al., 2020). In addition, a method was proposed to offer a cascaded estimation approach for accurate forecasts of construction costs at the conceptual phases of building projects using linear regression and artificial neural networks (Dursun and Stoy, 2016).
Multi-sensor based landslide monitoring via transfer learning
Published in Journal of Quality Technology, 2021
Wendong Li, Fugee Tsung, Zhenli Song, Ke Zhang, Dongdong Xiang
In multi-sensor based landslide monitoring, the essential tasks are the detection and early-warning of anomalies in auto-correlated data streams. Statistical process control (SPC) charts are important tools in such situations, and find broad applications in a diverse range of fields (Shu and Tsung 2003; Castagliola and Tsung 2005; Shang, Tsung, and Zou 2013; Shen et al. 2013). In the literature of SPC, the online monitoring of auto-correlated data streams and processes has been intensively studied. A commonly used strategy for univariate process is to fit a time series model and then apply a control chart to the residuals, while vector autoregression (VAR, cf., e.g. Nicholson et al. 2020) has been widely used in multivariate time-series modeling and monitoring. So far, many SPC methods have been proposed for monitoring univariate/multivariate auto-correlated processes. See, for instance, Apley and Tsung (2002); Shu, Apley, and Tsung (2002); Apley and Lee (2008); Capizzi and Masarotto (2008); Guo, Paynabar, and Jin (2012); Li and Qiu (2020); Qiu, Li, and Li (2020); Xue and Qiu (2020); Yu, Wu, and Tsung (2021) and the references therein.
Bayesian online robust parameter design for correlated multiple responses
Published in Quality Technology & Quantitative Management, 2021
Shijuan Yang, Jianjun Wang, Xiaolei Ren, Tingyu Gao
This paper assumes that the noise factor(s) can be observed in real-time during the production phase. The noise factor observations collected orderly at equally spaced time intervals can be regarded as time series data. We study the dynamic fluctuation mechanism of noise factors by fitting time series models to these data. The autoregression moving average model (ARMA) is commonly used to describe the dynamic fluctuation laws of univariate stationary series. When the series is nonstationary, the autoregression integrated moving average model (ARIMA) can be used (Yuan et al., 2016). Both the ARMA model and the ARIMA model can be subdivided into three categories, they are AR model, MA model, and ARMA model (or ARI model, IMA model, ARIMA model). Since both ARMA, ARIMA, MA, and IMA models involve the weighted sum of the current and past values of a white noise series, their predictive performance is not as good as that of the AR or ARI model (Chatfield, 2016). Therefore, only the AR and ARI models are considered in this paper. When there are multiple correlated noise factors in the system, the vector autoregression model (VAR) can be used to describe the autocorrelation of each noise factor and the correlation between multiple noise factors.
Lithium industry in the behavior of the mergers and acquisitions in the US oil and gas industry
Published in Energy Sources, Part B: Economics, Planning, and Policy, 2018
Manuel Monge, Luis A. Gil-Alana
We use wavelet analysis (Aguiar-Conraria and Soares 2011a, 2011b)1https://sites.google.com/site/aguiarconraria/joanasoares-wavelets. to detect the evolution in time frequencies, paying particular attention to the trend or long-run component in the time series (low frequency) and the seasonality or the short-run component and the rapid changes in the time series (high frequency). The authors focus on dynamic correlations based on wavelet coherence between lithium industry, represented by Solactive Global Lithium Index, and M&A in the US O&G industry using monthly data. The evolution of these correlations in time as well as for different frequencies has been analyzed. In addition, the wavelet phase-difference has been analyzed. This approach distinguishes between different behaviors with different horizons. Finally, in relation with causality and wavelets, Olayeni (2016) argues that measuring causal effects using continuous wavelet transform (CWT) has been particularly problematic because such measures as wavelet coherence only embody amplitude between the variables; the information on the direction necessary for scooping out causal links is unavailable. However, the useful information on lead–lag relationships is encoded in the phase-difference. In addition, Dhamala et al. (2008), who try to undertake causality in non-parametrical wavelets, mention that the trouble lies in computing the spectral matrix factors in order to derive the minimum phase. This process involves inverse Fourier to communicate between the time and frequency domains. For this reason, Granger causality test after vector autoregression (VAR) model estimation has been used to examine the causality direction between both time series.