Prediction Liquidated Damages via Ensemble Machine Learning Model: Towards Sustainable Highway Construction Projects

Sustainability, Journal Year: 2022, Volume and Issue: 14(15), P. 9303 - 9303

Published: July 29, 2022

Highway construction projects are important for financial and social development in the United States. Such types of usually accompanied by delay, causing liquidated damages (LDs) as a contractual provision vital agreements. Accurate quantification LDs is essential contract parties to avoid legal disputes unfair provisions due lack appropriate documentation. This paper effort sought develop an ensemble machine learning technique (EMLT) that combines algorithms Extreme Gradient Boosting (XGBoost), Categorical (CatBoost), k-Nearest Neighbor (kNN), Light Machine (LightGBM), Artificial Neural Network (ANN), Decision Tree (DT) prediction highway projects. Key attributes identified examined predict interrelated correlations among influential features accurate forecast models assess impact each delay factor. Various machine-learning-based were developed, where different modeling outputs analyzed compared. Four performance matrices such Root Mean Square Error (RMSE), Absolute (MAE), Percentage (MAPE), Coefficient Determination (R2) used evaluate accuracy implemented (ML) algorithms. The implied developed EMLT model has shown better compared other ML-based models, it highest 0.997, DT, kNN, CatBoost, XGBoost, LightGBM, ANN with 0.989, 0.988, 0.986, 0.975, 0.873, 0.689, respectively. Thus, findings this research designate can be effective administrative decision adding tool forecasting LDs. As result, emphasizes ML’s potential aid advancement computerization comprehensible subject investigation within building

Language: Английский

---

Blockchain-Based Information Supervision Model for Rice Supply Chains

Computational Intelligence and Neuroscience, Journal Year: 2022, Volume and Issue: 2022, P. 1 - 17

Published: March 29, 2022

Rice is a major food crop around the world, and its various quality safety problems are closely related to human health. As an important area of research, rice supply chain has attracted increasing attention. Based on blockchain technology, this study investigated data privacy circulation efficiency caused by complex networks, long cycles, risk factors in each link. First, we deconstructed link at information level established key classification table for On that basis, built supervision model based blockchain. Various encryption algorithms used secure sensitive enterprises meet regulators' needs efficient supervision. Moreover, propose practical Byzantine fault-tolerant consensus algorithm scores credit enterprise nodes, optimizes selection strategy master ensures high low cost. Then, prototype system open-source framework hyperledger fabric, analyzed model's viability, implemented using cases. The results indicated proposed can optimize process regulators provide feasible solution grain oil.

Language: Английский

---

SAC-ConvLSTM: A novel spatio-temporal deep learning-based approach for a short term power load forecasting

Expert Systems with Applications, Journal Year: 2023, Volume and Issue: 237, P. 121487 - 121487

Published: Sept. 15, 2023

Language: Английский

---

Time Series Prediction in Industry 4.0: A Comprehensive Review and Prospects for Future Advancements

Applied Sciences, Journal Year: 2023, Volume and Issue: 13(22), P. 12374 - 12374

Published: Nov. 15, 2023

Time series prediction stands at the forefront of fourth industrial revolution (Industry 4.0), offering a crucial analytical tool for vast data streams generated by modern processes. This literature review systematically consolidates existing research on predictive analysis time within framework Industry 4.0, illustrating its critical role in enhancing operational foresight and strategic planning. Tracing evolution from first to revolution, paper delineates how each phase has incrementally set stage today’s data-centric manufacturing paradigms. It critically examines emergent technologies such as Internet things (IoT), artificial intelligence (AI), cloud computing, big analytics converge context 4.0 transform into actionable insights. Specifically, explores applications maintenance, production optimization, sales forecasting, anomaly detection, underscoring transformative impact accurate forecasting operations. The culminates call action dissemination management these technologies, proposing pathway leveraging drive societal economic advancement. Serving foundational compendium, this article aims inform guide ongoing practice intersection 4.0.

Language: Английский

---

Parameter Adaptive Non-Model-Based State Estimation Combining Attention Mechanism and LSTM

IECE transactions on intelligent systematics., Journal Year: 2024, Volume and Issue: 1(1), P. 40 - 48

Published: May 29, 2024

Nowadays, state estimation is widely used in fields such as autonomous driving and drone navigation. However, practical applications, it difficult to obtain accurate target motion models noise covariance.This leads a decrease the accuracy of traditional Kalman filters. To address this issue, paper proposes an adaptive model free method based on attention parameter learning module. This combines Transformer's encoder with Long Short Term Memory Network (LSTM), obtains system's operational characteristics through offline measurement data without modeling system dynamics characteristics. In addition, output module, expectation maximization (EM) algorithm estimate parameters online, filter estimation. was validated using GPS trajectory path dataset, experimental results showed that proposed has better than other models, providing effective for deep networks

Language: Английский

---

Parking demand forecasting based on improved complete ensemble empirical mode decomposition and GRU model

Engineering Applications of Artificial Intelligence, Journal Year: 2022, Volume and Issue: 119, P. 105717 - 105717

Published: Dec. 17, 2022

Language: Английский

---

TreeCN: Time Series Prediction With the Tree Convolutional Network for Traffic Prediction

IEEE Transactions on Intelligent Transportation Systems, Journal Year: 2023, Volume and Issue: 25(5), P. 3751 - 3766

Published: Oct. 27, 2023

The complexity of traffic scenarios, the spatial-temporal feature correlations pose higher challenges for prediction research. Traffic model is an essential method in this research field, primarily focusing on capturing features among nodes and their neighboring nodes. However, existing methods lack comprehensive consideration directional hierarchical They are mostly applicable to scenarios with random uniform distribution nodes, but not suitable more complex small-scale aggregation scenarios. Therefore, study proposes Tree Convolutional Network (TreeCN), a tree-based structure. data design TreeCN focus relationships represented by plane tree matrix constructed as spatial matrix. TreeCN, full convolution network, performs bottom-up structure complete task node capturing. In study, thoroughly compared statistical, machine learning, deep learning time series prediction. experimental results show that only well also exhibits outstanding effect distribution. Moreover, adheres principles Graph Networks (GCN) can further capture them. This expected make new handle improve accuracy.

Language: Английский

---

Attention based long-term air temperature forecasting network: ALTF Net

Knowledge-Based Systems, Journal Year: 2022, Volume and Issue: 252, P. 109442 - 109442

Published: July 17, 2022

Language: Английский

---

A novel intuitionistic fuzzy time series prediction model with cascaded structure for financial time series

Expert Systems with Applications, Journal Year: 2022, Volume and Issue: 215, P. 119336 - 119336

Published: Nov. 24, 2022

Language: Английский

---

Overview of Data Augmentation Techniques in Time Series Analysis

International Journal of Advanced Computer Science and Applications, Journal Year: 2024, Volume and Issue: 15(1)

Published: Jan. 1, 2024

Time series data analysis is vital in numerous fields, driven by advancements deep learning and machine learning. This paper presents a comprehensive overview of augmentation techniques time analysis, with specific focus on their applications within We commence systematic methodology for literature selection, curating 757 articles from prominent databases. Subsequent sections delve into various techniques, encompassing traditional approaches like interpolation advanced methods Synthetic Data Generation, Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs). These address complexities inherent data. Moreover, we scrutinize limitations, including computational costs overfitting risks. However, it's essential to note that our does not end limitations. also comprehensively analyzed the advantages applicability under consideration. holistic evaluation allows us provide balanced perspective. In summary, this illuminates augmentation's role machine-learning contexts. It provides valuable insights researchers practitioners, advancing these fields charting paths future exploration.

Language: Английский

---